Scientists often struggle to communicate their work, said Ralph Cicerone, president of the National Academy of Sciences, in his introductory remarks on the second day of the Arthur M. Sackler colloquium on The Science of Science Communication II. Even clear descriptions of evidence based on careful experiments, observations, or calculations do not always get through to many audiences. As a result, scientists have realized that they need to learn new ways of communicating scientific information to nonscientists.
In recent years the National Academy of Sciences, National Academy of Engineering, Institute of Medicine, and National Research Council have grown increasingly concerned about the communication of science-related issues to the public. As a consequence, the institutions have been inviting social scientists to meetings to learn more about the challenges of science communication. The institutions have learned about how people form beliefs and attitudes and why scientists sometimes get caught in the middle of political, economic, or moral disputes. They have learned about the economic and social factors that can shape science communications and about the potential of social networks. Those interactions led to a more systematic engagement when the Academy hosted the first Sackler colloquium on the Science of Science Communication in May 2012.
Like the first colloquium, the second was designed to push people in “uncomfortable ways,” said Cicerone. Social scientists were asked to bridge the gap from their controlled studies to the complex world in which communication actually takes place. Science communicators were asked to
augment their own professional judgment with scientific evidence about how they communicate. Subject-matter experts were asked to listen both to social scientists and to practitioners while making certain that they get the facts right. The object of the colloquium was to foster innovative thinking and fruitful new collaborations through interactions that would not have occurred otherwise.
Kathleen Hall Jamieson, Elizabeth Ware Packard Professor of Communication at the University of Pennsylvania’s Annenberg School for Communication, began her keynote address on the second day of the colloquium by looking at two broad communities that are involved in science communication.
The scientific community is part of an expert or elite community responsible for knowledge generation. This community also includes institutional entities such as the Bureau of Labor Statistics and the Congressional Budget Office that use the best available methods to generate knowledge, internally critique their methods to improve them, and police what it is that they communicate. This community seeks to make sure that it does the best it can to communicate what is knowable with the best available evidence. It usually does not achieve complete certainty, but it also seeks to communicate the levels of uncertainty associated with knowledge.
Journalists are part of a community that does not generate knowledge; rather, it uncovers and transfers knowledge that already exists. Journalism, too, is responsible for being transparent, for disclosing how it does what it does, and for policing itself. When it makes mistakes, the journalistic community, like the scholarly community, is expected to correct the record.
Both communities, when they perform their functions well, inform the policy-making community. In the process, they are able to hold the policy-making community accountable for its actions in relationship to the knowable. This model is not a completely accurate description of reality, Jamieson acknowledged, but it provides a framework for analysis and discussion.
The expert community’s policing process requires that the public understand how it knows what it knows. When the expert community certifies that it knows something, the public should be confident that it does.
Sometimes the scientific community gets something wrong. For example, the peer-reviewed 1998 article by Andrew Wakefield that associated vaccination in children with a pervasive developmental disorder was a highly consequential case in which peer review failed and the expert community communicated something that it should not have communicated. The journalistic community also was at fault in this case by not uncovering the overall scientific consensus. Most journalists ran with the story, but one journalist began to investigate improprieties in the research that eventually led to its being discredited. When the scientific community learned of flaws in the evidentiary chain, it retracted the article. In this case, both communities eventually acted as they should have. People were still hurt, but fewer people were hurt than would have been if the two communities had not eventually maintained their standards of self-correction.
One problem with self-correction is that it can fan the suspicions of those who believe that these communities are inept, duplicitous, or partisan, Jamieson observed. To prevent such suspicions from undermining the integrity of these communities, they need to frame the correction in such a way that the public understands that these communities are acting as they should. If corrections were common, the functioning of the communities could be questioned, but most of the time they get most of what they do right. Some percentage of the population will never be persuaded on particular points, Jamieson acknowledged. But democratic systems have mechanisms to decide when sufficient numbers of voters agree to take action.
Factors that Undermine Credibility
However, this model of knowledge generation and dissemination can be undermined, Jamieson continued.
First, the custodians of knowledge can be challenged. For example, nonpartisan institutions such as the Congressional Budget Office can be attacked as not using the best available evidence and methods but instead articulating a partisan position. Similarly, politicians may attack journalists as partisan when reporting infringes on their ability to construct a reality for the electorate.
Second, individuals have a tendency to see evidence through a partisan filter. When a partisan perspective is applied, evidence may be used in a partial fashion, not in a way that represents the best use of all the available evidence.
Effectively responding to such attacks on the use of the best available evidence requires acknowledging the two personae functioning in every communication, said Jamieson. The first is the communicator; the second
is the intended audience. Both have obligations that they bring to any exchange.
The scientific community needs to be credible, impartial, and respectful. Polemics are outside the engagement process that is appropriate either for journalists or for the expert community. The scientific community therefore needs to view its audience as intelligent, thoughtful, and worthy of engagement. Scientists do not have the responsibility to make policy decisions, but science has the potential to establish a context within which people can decide to act. It can lay out the consequences of doing something and the mechanisms that produce those consequences. In this way, it can enable people to make decisions based on the best knowledge backed by the best available evidence.
The scientific community also needs to find common ground with the audience it is trying to reach. Effective communication is built on shared premises or assumptions. An audience has to invest meaning in a communicative exchange. This meaning exists at the intersection of a text, a context, and a receiving audience.
Finally, scientists need to try to share knowledge, not impose it. They can inoculate against opposing claims by explaining what is known and what is not known.
The Attributes of Science Communications
The integrity of all evidence needs to be rigorously scrutinized. But science communication also needs a way to convey the existence of a consensus, Jamieson observed. Communications can emphasize that the experts who have formed the consensus have been right in the past. They need to employ a voice that is credible, impartial, and trustworthy. They need to communicate that scientists care about this issue and problem but also that they care about the integrity of their methods and in providing information that is as accurate as possible given the existing state of knowledge. Sources of funding need to be disclosed so that people can judge the effect that a funding source might have on the credibility of scientific results.
Journalists also have a responsibility of determining when a consensus does not exist. When journalists do not scrutinize the available evidence, they run the risk of conveying misinformation. Journalists can err both by failing to report on consensus and by assuming a consensus that does not in fact exist.
Besides conveying the presence or absence of consensus, science communications need to be nonpartisan, so that they can counter partisan filters. They need to provide a construction of reality that lets people
understand what is known and how it is known. And science communications need to use a voice that conveys respect for the audience.
Jamieson used several examples to illustrate her points, one of which involved the following 2013 headline from the New York Times:
“Arctic Ice Makes Comeback from Record Low, but Long-Term Decline May Continue”
This headline frames the issues discussed in the article in a highly misleading way, Jamieson contended. It fails to point out that the long-term trend in Arctic sea ice has been steadily downward. It cites further declines as possible but uncertain. As Jamieson said, an asteroid “may” destroy the National Academy of Sciences building, but it is highly unlikely.
On the Fox News website, this news was framed as follows:
“Arctic sea ice up 60 percent in 2013”
The article went on to note that the increase is “a dramatic deviation from predictions of ‘an ice-free Arctic in 2013.’” The article also said, “The surge in Arctic ice is a dramatic change from last year’s record-setting lows, which fueled dire predictions of an imminent ice-free summer.”
However, the article also cited data from the National Snow and Ice Data Center of Boulder, Colorado, showing the overall trendline in sea ice as going down. This point of common ground can be used to forward, deeply and respectfully, an alternative argument—that the amount of sea ice has been declining over time. Furthermore, this argument can be supported by visual demonstrations and additional data to reinforce the concept that the trendline is down. In addition, vivid images, clarifying metaphors, and evocative narratives can be used to reinforce the overall message—such as the idea that the reduction of sea ice is dramatic enough to see from the moon, or reports of changes in polar bear behavior and mortality.
Conveying a consensus to the public requires time, education, and breaking through partisan filters, Jamieson concluded. When scientists are positioned as nonpartisan, evidence is more likely to be heard and scientists are less likely to be seen as polemicists or persuaders.
For some important issues in science and technology, public engagement and communication need to move upstream to much earlier in the research and development process, said Nick Pidgeon, professor of environmental psychology at Cardiff University in Wales. The objective is not just to persuade someone that a new technology should be accepted. Rather, a more discursive and two-way approach to public engagement can foster better overall decision making.
In the case of nanotechnology, for example, the Royal Society report Nanoscience and Nanotechnologies: Opportunities and Uncertainties (Royal Academy of Engineering, 2004) made the point that early engagement and dialogue can achieve several critical ends:
• Incorporating public values in decisions,
• Improving decision quality,
• Resolving conflict,
• Establishing trust and legitimacy, and
• Education and information.
Upstream engagement also has several obvious difficulties, Pidgeon continued. People are likely to know less about a technology at an early stage of research. Mental models of risk processes are likely to be absent or ill formed, with analogies instead serving as a proxy of risk. Both the future course of the science and potential regulatory needs or gaps will probably be uncertain, and the promoters and detractors of a technology are likely to issue both hype and dystopian narratives.
All of these difficulties are apparent when considering geoengineering—the intentional manipulation of the Earth’s climate to counteract warming or other aspects of climate change. Modification of the climate has been discussed for decades, but geoengineering as a way to counter climate change has been seriously discussed for only a few years, and scientists are deeply conflicted about it. It involves reflecting solar radiation back into space to lower global temperatures or removing carbon dioxide from the atmosphere, either of which would require engineering projects of immense scope. Neither the feasibility nor the full consequences of such methods for geoengineering are yet known.
In 2010 the equivalent in the United Kingdom of the National Science Foundation (NSF) commissioned the Stratospheric Particle Injection for Climate Engineering (SPICE) project. Conducted by a large consortium of
engineers and scientists, the project explored the possibility of delivering reflecting aerosols through a 20-kilometer pipe tethered to a giant weather balloon. The project involved laboratory experimentation, modeling, and background review of the project and its possible impacts and risks. The project also included a proposed field trial—a small-scale 1-kilometer mock-up with a small balloon spraying water to answer some basic engineering questions.
The project was approved by two university ethics boards on the grounds that it did not jeopardize human health, interfere with animals, or have any detrimental effect on the environment. However, the reaction of the press and of some nongovernmental organizations was intense. Given the sensitivities of the technology and its implications, the research governance protocols that allowed it to be approved have to be questioned, Pidgeon said.
A Framework for Responsible Innovation
In response to the controversy, the SPICE researchers were asked to address five criteria before the pipe and balloon test could go ahead. One was to identify mechanisms to understand wider public and stakeholder views regarding envisaged applications and impacts of the experiment. Pidgeon’s team then was asked to design a protocol that would enable members of the public who knew very little about the technology to form a considered opinion on whether the field trial should go ahead.
Developing such a protocol was immensely challenging both in conception and in methodological terms, said Pidgeon. It required intensive piloting, extensive engagement with the SPICE team and other geoengineering experts, and input from a stakeholder advisory panel. Three two-day workshops were conducted in different British cities, with 10 members of the general public selected to participate in each workshop. The aim in each workshop was to bring participants up to speed on the science and ethics of geoengineering and then to solicit their views. Methodological considerations in holding the workshops included framing the issue and the materials and experts to be employed. For example, workshop participants needed to be exposed to different framings of the issues involved (on both technical and ethical questions) to avoid presupposing their positions.
A particular methodological consideration was which people should be included in the workshops. There is a difference between an audience chosen essentially at random, such as the jury approach eventually adopted in this case, and an audience with a preexisting interest in a question. People with a preexisting interest can have a different set of attitudes prior to an engagement, yet they can be just as important in deciding an
outcome as people without a preexisting set of attitudes. Another question for future dialogues around geoengineering involves whether participants should be from developed countries or from the less developed countries that are likely to suffer the most severe immediate consequences of climate change or unintended adverse impacts of geoengineering. These and other questions regarding the make-up of participants in such exercises are still being debated.
Following the workshops, very few participants wanted to rule out the 1-kilometer test, Pidgeon reported. They felt that it was a good thing for scientists to explore the topic, even though their views on the use of stratospheric aerosols were very negative, since people are disturbed by the thought of interfering with natural systems on a planetary scale.
In controversial areas, Pidgeon concluded, scientists and science communicators need to respect the views of the public if science is to progress. In addition, decision making over issues that will affect our lives in the future requires an emotional commitment as well as the analytical weighing of costs and benefits.
Upstream dialogue is extraordinarily important, agreed William Hallman, professor and chair in the Department of Human Ecology at Rutgers University, in his comments on Pidgeon’s presentation. But scientists and science communicators need to be very careful with what they do upstream, “because what we put in the water upstream we end up drinking downstream.”
One issue is that consensus needs to exist that a particular problem is worth discussing, Hallman said. Once this consensus exists, discussion can proceed on whether a particular technology is the right way to solve a problem.
Upstream dialogue is also complicated by the fact that most members of the public will know very little about the topic (though this often will not prevent them from stating an opinion). Those initiating the dialogue therefore need to be very careful about what they bring to the discussion.
Rick Borchelt, director of communications and public affairs for the Department of Energy’s Office of Science, also agreed that upstream engagement is laudable as a democratic ideal. In practice, however, it is fraught with potential problems.
Climate change is distinct in posing a dire problem that needs to be solved. But other technologies do not necessarily present problems in their
earliest stages of development. The question therefore becomes whether upstream engagement is generally applicable to many of the issues that interest science communicators.
In addition, upstream engagement runs the risk of inciting others to develop counternarratives that might not have existed if the engagement were not performed. If the objective of upstream engagement is to fend off future controversy, the question becomes whether the efforts should be done without reference to the possible controversy or as a fully transparent exercise. One possibility is that engagement will create an arms race downstream over issues. In general, the upstream environment is rarely free of the controversies that upstream engagement presupposes, Borchelt said.
In general, the nature of the engagement process is critical. For example, scientists need to both provide information and gather information through listening. If they are not prepared to listen as well as talk, they should not be going into an engagement opportunity.
For a guideline to change behavior, it has to be memorable and actionable, said Rebecca Ratner, professor of marketing at the Robert H. Smith School of Business at the University of Maryland. An exception involves guidelines so complicated that they cannot be easily remembered, in which case a checklist can be an effective way to influence behavior. But in general, a guideline must be remembered by the target audience, and they must be able to do what it recommends.
Nutrition is an area where guidelines are useful since people generally cannot consult a guide every time they make a decision about what to eat. For example, the food pyramid, which was developed in 1992, called for people to eat 6 to 11 servings of bread, cereal, rice, or pasta each day; three to five servings of vegetables; two to four servings of fruit; two to three servings of milk, yogurt, or cheese; and two to three servings of meat, poultry, fish, dry beans, eggs, and nuts; in addition to eating fats, oils, and sweets sparingly. The guideline was memorable, Ratner said, though people had a hard time remembering the recommended numbers of servings of each food group. However, it was less actionable, because people were not sure about the size of a serving and it was hard to keep track of servings over the course of a day.
In 2005 the U.S. Department of Agriculture released a new guideline called MyPyramid consisting of five food groups: grains, vegetables, fruits, milk, and meat and beans. People were asked to go to a website where they would enter their age, sex, and how much exercise they got in a typical day, and the website would produce individualized guidelines
for how much people should eat, in ounces and cups, from each of the five groups.
Ratner and her colleague Jason Riis studied the memorability and actionability of the food pyramid and found it wanting in both dimensions. After studying the personalized guidelines in each of the five categories for as long as they wanted, people were asked immediately after to recall the five numbers they had just seen. Only 19 percent of participants correctly recalled the numbers in all five categories, and less than 1 percent correctly recalled the correct numbers in all five categories 1 month later. People also were unsure about the size of an ounce of food, and again they had difficulty tracking their consumption over the course of a day.
A Better Way
Porter Novelli, the public relations firm that helped to develop MyPyramid and the original food pyramid, was testing another message at about the time that the food pyramid was introduced: fill half your plate with fruits and vegetables at every meal. Ratner and Riis saw this as a much more memorable and actionable recommendation, and testing confirmed their hunch. Immediately after studying it, 85 percent of participants correctly recalled the guideline, and 62 percent recalled it one month later. The guideline was also actionable, since people can tell when roughly half of a plate is full of fruits and vegetables and they did not need to keep track of their consumption over the course of a day.
When a guideline is memorable and actionable, people are more motivated to follow the guideline, Ratner stated. For example, in a comparison of the MyPyramid and the half-plate recommendations, both dieters and nondieters demonstrated more interest in adhering to the latter, with a particular increase in motivation among dieters.
Ratner said that she was delighted to learn that the new Obama administration intended to revamp the government’s messaging about nutrition. In 2011 a new guideline, ChooseMyPlate.gov (http://www.choosemyplate.gov), incorporated the half plate of fruits and vegetables into a much more schematic treatment of the five food groups. This revision “definitely has the potential to help people follow nutrition science,” she said.
Extensions Beyond Nutrition
In general, Ratner listed four attributes that make a message memorable and actionable, no matter what its subject.
First, it needs to be simple. Examples of simple messages are “got milk?” “drop and roll,” and “just do it.” They are easy to remember and easy to follow.
Second, the message needs to be easy to visualize. For example, the “got milk?” message is associated with a simple and easy-to-remember advertising image.
Third, a message should specify when to engage in an action. For example, when people were asked to take vitamins each day, those given an action plan for taking the vitamins—such as doing so every morning at breakfast—were nearly twice as likely to do so than those without an action plan.
Fourth, a message should embed a trigger to take action. For example, the dining hall message “live the healthy way, eat five fruits and veggies a day” was not nearly as effective as the message “each and every dining hall tray needs five fruits and veggies a day,” because the latter reminded people to eat fruits and vegetables—but only in dining halls that had trays.
What Should Be Simplified?
Besides being memorable and actionable, messages need to be motivational and plausible, said William Hallman, professor and chair in the Department of Human Ecology at Rutgers University, in his comments on Ratner’s presentation. For example, people need to understand why eating more fruits and vegetables is important. The graphic does not convey that information, so the reasons for eating some foods rather than others would need to be learned.
Also, not everything should be simplified. Much nutrition advice is simplified—“dangerously so,” Hallman said. “Sugar is death” is a simple message, but in purely biochemical terms it may be more accurate to say that no sugar is death. Messages can be simple, plausible, memorable, and actionable—“and just plain wrong.”
What Should Be Actionable?
Many science communications are not actionable, added Borchelt. Indeed, science communicators often would prefer that scientific information not be dragged into a political arena where it can be used to justify action of one kind of another.
Where action is desired, constraint recognition often stymies action. People may recognize that climate change is a dire problem, but they may also believe that nothing they do will make a difference.
People often do not have enough information to determine whether they have the ability to take action, Borchelt said. Simple and actionable
messages can help people choose between two evidence-based diets, but that is not the typical situation in science communication.
When Punam Anand Keller, the Charles Henry Jones Third Century Professor of Management at the Tuck School of Business at Dartmouth College, was a child, she had the job of offering drinks to people whom her father counseled at their home, but many declined her offer. Finally her father asked her what question she was asking, and she replied, “Do you want a drink?” “Aha!” said her father. “That’s your problem! You gave them an alternative.” As soon as she began asking whether people wanted something hot or cold, more people accepted her offer.
Choice architecture, or the way options are designed and offered, can have a big influence on the decisions people make, Keller said. In particular, choice architecture is a useful communication method to simplify trade-offs when decisions have to be made by the audience. For example, simply telling someone to do something is not a very effective prompt for problem recognition. People are told to lose weight, save energy, volunteer, save money, and drink responsibly, but few do all five. Communication fails when it does not connect a person with a problem, a question, a goal, or a dream. It also fails when people are not motivated to take action or make a judgment based on the information conveyed in the communication. In contrast, messages that prompt people to take action can be a very useful tool to increase the effectiveness of the communication.
Keller described four of the most common forms of choice architecture. The first is an opt-in approach, as when colloquium participants are told, “Check the box if you will attend the Sackler webinar on the Science of Communications.” The second is an opt-out approach, which would use the question “Check the box if you will not attend the Sackler webinar on the Science of Communications.”
The third option is an active choice, in which participants pick among options. For example: 1. “I will attend the Sackler webinar on the Science of Communications.” 2. “I will not attend the Sackler webinar on the Science of Communications.”
The fourth option is enhanced active choice, in which the choices specify the advantages of choosing that option. For example, colloquium participants might be asked the following:
1. I will attend the Sackler webinar because it is important for me to discover new communication ideas and share research with academics and practitioners.
2. I will not attend the Sackler webinar because other commitments prevent me from discovering new communication ideas and the opportunity to share research with academics and practitioners.
In creating enhanced active choices, those designing the choices obviously face the issue of how directive to be and whether the choices are effective, Keller observed. But it can be an effective tool in getting people to take beneficial actions. Enhanced active choice can be personal, motivating, and interactive, which can help engage someone who would otherwise be uninvolved.
As an example, Keller cited an enhanced active choice involving flu shots for hospital employees. The enhanced active choice asked the employees to check one of two boxes:
1. I want to remind myself to get a flu shot.
2. I want a reminder to get a flu shot.
Providing employees with an enhanced active choice was 50 percent more effective than the active choice of asking them to check either “I don’t want a reminder to get a flu shot” or “I want a reminder to get a flu shot.”
Another example involved enrollment for automatic refills of prescription drugs. When people were presented with a button on a website that they could click to enroll, about 12.5 percent did. But when a second button was added that said, “I prefer to order my own refills,” thereby forcing people to make a choice not to enroll, almost twice as many people enrolled. Comparable results are seen with mailed responses, Keller added, so the outcome is not dictated simply by forcing people to choose.
Finally, Keller described an experiment with voice recordings designed to convince people to get their medications through the mail. In an opt-in approach, the recording said:
“Would you like to speak with someone about getting started with mail service? Please say ‘yes’ or ‘no.’”
In the enhanced active choice, people were told:
If you would rather pay more and continue making many trips to the pharmacy, say 1. If you’re tired of paying more and making unneeded trips to the pharmacy, say 2.
The enhanced active choice was 66 percent more effective, and despite how directive the choices were, most respondents reported being satisfied with how the automated voice response portion of the call was handled.
Enhanced active choice can provide a greater sense of control and belonging, be simpler and more urgent, and convey a sense of trust and shared goals, Keller concluded. As such, it meets the objectives of self-enhancement and accuracy desired of all science communication and can help nudge people in a scientifically accepted direction.
The Ethics of Enhanced Active Choice
Enhanced active choice clearly raises ethical issues, said Hallman, in his comments on Keller’s talk. It is designed to persuade, not just to educate, which raises the question of whose interests are being served. After all, the standard question “Do you want fries with that?” could be made into an enhanced active choice question, which is not necessarily in a customer’s best interest. Hallman also asked for a third choice for any such question: 3. “Please treat me as an adult and stop patronizing me because I can make my own decisions.”
That said, Hallman added, people who most need to understand complex information are often in the worst position to understand it. Simplifying and creating choices that make sense for individuals and for society can, in these circumstances, have many advantages.
Enhanced Active Choice to Serve Science
Like upstream engagement, enhanced active choice is a tool, noted Borchelt. For science communication, a relevant question is whether the enhancement can incorporate science. The enhancements are often dictated by people with a vested interest in promoting a behavior. Will these people use science to shape choices, or will they pursue a different agenda?
If science is used to create an enhanced active choice, a related question is whether that use will harm the credibility of science. A goal of science communication is to build trust in science and in scientists. Incorporating science into decision-making architectures could backfire if people feel they are being manipulated, Borchelt said.
Business communication, including marketing and public relations, differs from science communication, but it nevertheless can provide some valuable lessons to science communicators, said Davis Masten, former head of the design consulting company Cheskin, and Peter Zandan, global vice chair and worldwide research practice group leader at Hill+Knowlton Strategies, during a joint presentation on the second day of the colloquium.
“Science is moving like a freight train,” said Masten. Though people may know little about it, science is a major part of their lives. Yet, today, science communication is being outmaneuvered, and this outmaneuvering is going to increase. As Zandan pointed out, businesses are now spending more than $1 trillion to get out their messages, with about half that amount devoted to targeted marketing designed to reach individuals. Zandan was challenged to come up with an estimate for how much is spent on science communications, but it is probably less than a billion dollars (excluding education)—so less than a thousandth as much. In fact, businesses are spending $9.5 billion a year just to research the effectiveness of their messages.
Furthermore, many of the techniques businesses use to reach audiences and to assess the effectiveness of their messages are derived from the social sciences, Zandan continued. Ironically, the work of the scientific community has been adapted by business more than it has been by the scientific community.
Science is still respected much more than other professions, largely because of its devotion to integrity and truth. If science communicators could use this respect to amplify their messages, they could have a much greater impact than they do today.
New Technologies and Social Media
As Masten said, within a few years, more than five billion people will have smartphones worldwide. The average city now has a billion sensors in it, and the number will be 10 billion by 2020. This amount of power and connectivity could make science relevant to the choices people make every day, and science communicators could help make that possibility a reality.
Social media also have changed the nature of engagement, said Zandan. About a million people read the print version of the New York Times each day, but Facebook, LinkedIn, and other social media sites involve hundreds of millions of people. Furthermore, their use of these platforms generates data that can be used to increase and channel that use. Businesses are already using these data to look at the return on investment for their messaging. As Zandan said, “The effectiveness that these platforms have provided is truly transforming communications for business.”
With social media, communications are not as expensive as with the mass media. A video can go viral at very little cost. Intelligence, creativity, and social savvy may be needed to create a popular video, but the potential to do so is not limited to business. Moreover, with a billion people a day using social media, even a penny per day from each of them to support science communication would represent a large amount of money.
The Need for Transparency
Zandan said that the hottest word in business today is “transparency.” The social media have helped businesses realize that they need to be truthful and socially responsible because deviations from truthfulness will be played out in public.
In addition, everything that moves in business is being measured. Business is focused on its return on investment, but it no longer looks at this return purely in financial terms. The return is related to all the goals that a business is seeking, which includes financial calculations but also broader measures of accountability.
Science communicators could benefit by applying the same emphasis on metrics and accountability in their work. Research on the effectiveness of messaging in business is about 1 percent, Zandan noted. For science communication, he suggested that 3 to 5 percent be allocated to communication research. Furthermore, businesses are available to collaborate on this research, drawing on the research they have adopted from academia.
The Need for Partnerships
Science needs to maintain its independence and objectivity, but collaborations with business may be a way for science to be heard and embraced by the public. Without such partnerships, science will become a smaller voice among the American public. As Masten said, the trillion dollars spent by businesses on messaging generates a lot of noise, whereas today all science has is a whisper.
Companies, governments, and nongovernmental organizations have different missions and objectives than do scientists, but they generally want to do what is right, said Zandan. They are receptive to scientific evidence and curious, because they need to listen to survive. “They need you just as much as you need them.”
Partnerships can involve all sectors of society. For example, the Science and Entertainment Exchange at the National Academy of Sciences has led to more than 700 consultations between the scientific community and the people who make movies and television shows.
Another way to maintain trust and engagement is to reach out to the 20 million students enrolled in colleges and universities in the United States. In addition, museums receive almost one billion visits each year in the United States, providing another superb opportunity to engage the public in science.
Science serves the nation and the world, Zandan and Masten concluded. If science is called into question, a loss of trust could damage a national and international asset.
For decades, social science research has reflected a dichotomy between mass media that broadcast to large undifferentiating audiences and interpersonal communication among people talking with each other, observed Duncan Watts, principal researcher at Microsoft Research. But this traditional dichotomy has been dissolving. The mass media are fragmenting into different channels and platforms, facilitated by a surge in recommendation engines. At the same time, individuals are being empowered by new technologies to grow audiences that are in some cases as large as those of traditional network television. This nearly continuous distribution of production has led to the new concept of “mass personal communication.”
Even as technology and the media landscape have changed, the questions asked in the social sciences have remained remarkably stable. In the 1940s Harold Lasswell laid out the essential problem: “Who says what to whom, through which channel, and with what effect?” It is a straightforward question, said Watts, but answering it is extraordinarily difficult. Observing who says what to whom is hard to do at scale, and the difficulty is compounded by the multiplicity of channels. And measuring the effects of all this communication is even harder.
The Web 2.0 revolution may be bringing the answer within reach. To social scientists, social media platforms are like telescopes were to astronomers, said Watts. They are instruments that make the invisible visible and enable new kinds of science. Social scientists can now do the same kind of science that physicists and biologists do. In particular, Twitter is an almost ideal platform to address Lasswell’s question. Everyone from the president of the United States to private individuals communicating with their friends is on Twitter. They have a well-defined attention graph, because the only reason to follow someone on Twitter is to hear what that person has to say, which is different from a site like Facebook. Social scientists can track the diffusion of information using tweets containing shortened URLs, which encompasses not everything on Twitter but a lot. And a restricted version of influence can be measured by looking at retweets, click-throughs, and conversations.
Elite Users on Twitter
Watts described a research project in which he and his colleagues looked at everything on Twitter over an 8-month period, which they distilled down from 5 billion tweets to 260 million that contained bit.ly URLs. They then used Twitter Lists, which users employ to filter their streams, to produce what are essentially crowd-sourced labels for users named in the lists. For example, if someone is listed in thousands and thousands of lists that have the word “celebrity” in the title, that person is probably a celebrity.
Using this technique, they identified four classes of “elite” users: celebrities, people in the media, people in other kinds of organizations, and bloggers. They then ranked users by the frequency of being listed and measured the flow of URLs from the top 5,000 users in each category to the mass of Twitter users. Of the total number of URLs on a random user’s newsfeed, almost half come from one of these 20,000 people, which represents a tremendous concentration of attention on a network as distributed as Twitter.
With the exception of the other organizations category, the top users within each category pay more attention to other users in the category, a phenomenon known as homophily. This latter observations makes sense, said Watts, since organizations use Twitter not only to broadcast information but to hear what people are saying about them.
However, the picture is different for retweets. Celebrities and organizations do not retweet to each other much, but the media and especially bloggers retweet within their categories more than they do to the other categories.
These data also can be used to analyze the flow of information through the network. Research done since the 1950s has demonstrated that trusted ties are more important than the mass media in determining individual opinions. But not all people are equally influential. A category of people called opinion leaders act as filters between the mass media and the masses. These opinion leaders absorb what is happening in the media, decide what is interesting, and pass that content to other people.
This effect can be quantified using Twitter data. Almost half of media-originating URLs are received from people who can be identified as opinion leaders. These opinion leaders consume more media content than average users, are more active on Twitter, and have more followers. They also are followers, in that they receive much of their information indirectly. In the past, opinion leaders have been treated as “influencers” who act as mini-broadcast stations, and businesses have devoted substantial effort
to identifying these influencers to use them as conduits for marketing. As one observer wrote, “Influencers have become the ‘holy grail’ for today’s marketers.”
This is an appropriate metaphor, said Watts, because the point of the Holy Grail is that it is never found. For example, after a video has gone viral, people can always come up with explanations for its success. But these explanations have not helped marketers or anyone else predict which videos are going to become popular, and the same observation applies to influencers.
The Twitter data substantiate this conclusion. Watts and his colleagues examined 74 million cascades where people retweeted messages containing bit.ly URLs, looking at the attributes of both the retweeters and the content of the tweets. Among the tiny percentage of retweets that traveled more than one or two hops from their source, the only attributes that made a difference were past local influence and the number of followers, and even those factors have only a small explanatory effect. All the other attributes measured had no effect on predicting which tweets would be widely disseminated, including how many people someone is following, how much they tweet, and when they joined. Surprisingly, the content of the tweet also does not influence the amount of dissemination.
Given the randomness with which tweets are retweeted, individual influencers are essentially impossible to identify. Instead, larger numbers of people need to be targeted to reach influencers. The only possible exceptions are individuals sharing the two attributes that make a difference in retweeting, who tend to be well-known individuals such as the President and celebrities or organizations such as the Weather Channel. But even given the possible advantages of targeting these individuals, a campaign to disseminate a message generally cannot ignore the large number of “ordinary influencers” with few followers.
Computational Social Science
The attributes of Twitter users may explain only a small part of how extensively their tweets are retweeted, said Noshir Contractor, Jane S. and William J. White Professor of Behavioral Sciences at Northwestern University and one of two commentators for the session, but that is not necessarily a bad thing. A world in which the social sciences could explain a substantial part of human behavior might not be a very enjoyable world.
One thing that Watts’s research demonstrates is the value of computational social science, Contractor said. Social science typically has relied on surveys, experiments, interviews, and ethnographies, but now it can test theories at scale using data provided by social media. However, it is
important to remember that this research involves only Twitter, and social science should not be reduced to Twitterology.
The Chaotic Evolution of Technology
Human desires and needs, including the need to communicate, remain the same over time, said Xeni Jardin, the founding partner and co-editor of Boing Boing, but the systems used to communicate have been changing very rapidly in the past few decades. In addition, platforms like Twitter evolve over time because their designers cannot predict how humans will use them. “That chaotic effect is part of what keeps me fascinated with technology,” Jardin said.
As an example of how social networks can influence people’s lives, Jardin described her experience with breast cancer. She was diagnosed in December 2011 and completed her primary treatment in September 2012. She accidentally live-tweeted her diagnosis, because she thought that she was a young and aggressively health-conscious person who could never get cancer and that the experience of having a mammogram would interest her tens of thousands of followers. Once she tweeted her diagnosis, she remained active on Twitter and other social networks, including her blog, throughout her treatment. She also became aware of a group on Twitter called #BCSM, which stands for breast cancer social media. Though she never went to a real-life support group, she was very active with a group of breast cancer patients and care providers through #BCSM.
This experience was “one of the most formative experiences of my life,” said Jardin. It was a revelation about how up-to-the-minute and critically important science information can be disseminated and shared in a meaningful way through Twitter, in some cases even before care providers get the information. This particular group and network would be a valuable case study for social scientists to examine the real-world consequences of social networks, Jardin said.
According to Deb Roy, professor at MIT and chief media scientist at Twitter, television and Twitter are intersecting to create a new hybrid form of communication. This new hybrid is an audience-driven movement in which people have chosen Twitter as a natural way to talk to each other while they watch television. In turn, the combination is driving change both in the television industry and in Twitter.
Television delivers a synchronous experience to a large number of people, as do many other experiences. For example, everyone in a given area experiences a sunset at the same time, and this experience can be
quite different when someone is with another person rather than being alone. Two people can look at a sunset together and talk about it, which can completely change the primary experience, Roy noted.
Twitter is public, so that tweets can flow freely within the network but also pop up on the front page of the New York Times. It also is a fast medium, where the chances of someone reading a tweet drop off rapidly with time. It can be live when people come together at a scheduled time to communicate. This enables it to function something like a soundtrack to a movie; Twitter exchanges often take the form of a social soundtrack around life in the moment.
Combining these two observations, Roy pointed out that the number of tweets containing the word “sunset” peaks at the same time as the sunset around the world. People see the sunset and tweet about it, making a solitary experience a shared experience.
The same phenomenon occurs with television shows. During the presidential debates between Barack Obama and Mitt Romney, memorable phrases, such as “binders full of women,” immediately spread through Twitter. “I didn’t have to wait for the media pundits,” said Roy. “My own network plus some people I never met before but who entered the conversation in the moment and created that dynamic social network around this event influenced how I encoded what just happened.”
The same thing is happening with sports, drama shows, and even advertisements. Roy has been involved in research that tracks the content of tweets commenting on a particular television show or advertisement. The result is a graph containing detailed information about the cross influences of connected audiences and content. These tweets happen both immediately after an event occurs and for a period of time afterward, as people discuss what they saw and what others said about it. They demonstrate both the social amplification of an observation over time and the scale on which such amplification occurs. In fact, said Roy, the Nielsen company is incorporating tweets into a new way of rating the viewership of television shows in the United States.
Promoting Science Through Twitter
Media companies have started using this phenomenon as a way of reaching potential viewers. For example, ESPN could insert a small clip from a basketball game into tweets mentioning the game. Someone watching the game could therefore be enticed to tune in when they received the
tweet. In this way, observed Roy, a mass media experience can be delivered in a targeted fashion to a particular audience using social media.
The same process could be used for science communication, said Roy. Both good science and bad science are mixed into television shows. Either kind of science message could provoke a conversation on Twitter that could help disseminate information about science. Furthermore, some mass events have prominent science components, such as earthquakes, which could be used to disseminate science information, such as how to prepare for natural disasters.
Providing a context that makes a science event relevant to an audience often transcends Twitter or any other single medium, Roy acknowledged. But Twitter has the potential to create connections that did not exist before. It can have a profound impact on how primary experiences are interpreted—whether the experience is a sunset, a television show, or science.
Other hybrid media with symbiotic relationships have existed in the past, said Contractor in his comments on Roy’s presentation. For example, when television became widely prevalent, TV Guide became the country’s most popular print magazine. Similarly, in the early days of the Internet, people who watched soap operas engaged in extensive conversations about the shows on Usenet sites. In fact, said Contractor, people who did not watch the shows got so caught up in the Usenet conversations that they began watching the soap operas being discussed.
He also observed that television broadcasts can be accompanied by not only second screens but third, fourth, and fifth screens. People with particular interests—such as science—could discuss one aspect of a show, while people with other interests discuss other aspects. This might offer a way to galvanize a segment of the public that is more interested in science than other parts of the public.
One big difference between social media today and the Usenet conversations, said Jardin, is that today’s social platforms are privately owned spaces. Perhaps something is being lost by not having a public forum in which these conversations can occur.
When Katherine Milkman, the James G. Campbell Assistant Professor of Operations and Information Management at the Wharton School of the University of Pennsylvania, was in graduate school, she became curious
about the list of most widely e-mailed articles on the New York Times website. Why did some articles make it onto the list while others failed?
To satisfy her curiosity, she had a web crawler built that visited the paper’s website every 15 minutes and compiled the locations and full text of all of the news articles on the site as well as whether they were on the most e-mailed list. Over about 3 months she collected data on more than 7,000 articles, which she and a coauthor then analyzed to determine what factors determine which articles make the most e-mailed list.
Hypotheses to Test
Several ideas from the social sciences informed the analysis. First, people care deeply about the impressions they make on others. This would suggest that people would be more likely to share interesting, surprising, useful, or positive news to increase their self-enhancement.
Another motive might be to increase social bonding through the sharing of news. In particular, sharing emotional experiences can bring people closer together, which would suggest that articles evoking strong emotions might be shared.
Sharing also could be a form of emotional regulation. Some stories can provoke activating emotions such as fear or awe that people might share with others as a way of making sense of those emotions. Other stories might produce deactivating emotions such as sadness that would cause people to withdraw into themselves. In this case, stories that produce deactivated emotions would reduce sharing.
Factors That Increase Sharing
The analysis showed, first, that the position of a story in the newspaper matters. A 1-standard-deviation increase in the time a story spends as the lead article on the New York Times homepage increased its likelihood of making the most e-mailed list by about 20 percent.
As hypothesized, more interesting, surprising, and useful articles were more likely to make the list. More emotional stories were also significantly more likely to make the list, particularly stories containing more activating emotions as opposed to stories containing deactivating emotions.
Translating the Findings to Science
To explore the question of how these findings translate to the sharing of science, Milkman and her colleague Jonah Berger gathered data reported for the first time at the colloquium. They asked approximately 4,000 authors of articles that were published in leading science and social
science journals in the first half of 2013 if they would provide lay summaries of their scientific discoveries. About 20 percent agreed, resulting in 845 summaries of new scientific discoveries. They then recruited a separate panel of 8,000 nonscientists to rate a randomly selected scientific summary, and they averaged these ratings to create a measure of how likely an article is to be widely shared.
The results show that findings published in psychology journals are the most likely to be widely shared, followed by economics journals, sociology journals, and, finally, science journals. Digging deeper into the data, articles about business, psychology, other social sciences, and mathematics are most likely to be widely shared, while discoveries in chemistry, human services, biochemistry, genetics, and ecology are least likely to be shared.
They then used an automated linguistic classification software program to count the frequency with which people were mentioned in each summary. More frequent references to humans dramatically increase the likelihood of science being shared. “We care about science about people,” said Milkman.
Surprisingly, after controlling for the disciplinary affiliation of an author, summaries written by women are significantly more likely to be shared than summaries penned by men, which was also true of articles written by women in the New York Times. But when men and women were coauthors of a scientific article and each wrote a summary of the findings, the summaries were equally compelling. So, women are choosing to work on more sharable topics than men, but they are no better than men at describing the same findings.
Comparing summaries written by coauthors of the same article revealed several key features that increased shareability. For example, one scientist wrote of a finding:
We’re trying to build new types of crystal by combining layers from different materials. We’ve previously shown these can have many applications in digital and analog electronics. In this work we were able to turn light into electricity with a high conversion rate using our new structures made from graphene and tungsten disulfide, both atomically thin layered crystals.
A coauthor’s description had a much greater likelihood of being shared:
We produced a device that, although atomically thin, can strongly absorb light and convert it to electricity in a very efficient way. For every 100 photons of light, 30 are converted to electricity, which is a value comparable to the best solar cells in the market.
The features that make a scientific summary more shareable are very similar to those at work in the New York Times, said Milkman. If a summary is more interesting or more likely to reflect positively on a sender, it is more likely to be shared. In addition, summaries that explain why a result is more useful or are more emotionally resonant dramatically increase the likelihood of sharing. Being more positive also has a significant though small effect on shareability.
Interestingly, men and racial minorities are more likely to say that they would pass along scientific summaries than women and Caucasians. These groups view a given summary as more interesting, emotion inducing, useful, and comprehensible, and in the case of minorities, more emotion inducing and likely to reflect positively on them if shared.
By choosing their words carefully, scientists can increase the likelihood that their discoveries will be widely shared, Milkman concluded. They should emphasize why their work is useful, rely on emotional language, emphasize the positive, and focus on what is interesting and surprising.
Making Science More Shareable
As Contractor pointed out in his comments on Milkman’s presentation, the features that make a story shareable do not have much to do with scientific information. He also pointed out that the discussion of sharing has been unidirectional. However, the public also has knowledge, insights, and data that it can usefully share with scientists, and listening to this input can help scientists convey their own insights more effectively.
Jardin observed that different kinds of emotions can generate shareability, from outrage over human exploitation to affection for kittens. She also pointed out that independent blogs have been under intense pressure to pander to clickability, linkability, and shareability. Science faces the challenge of becoming more shareable while staying true to the point of the work and not pandering to the lowest common denominator.
On that note, Watts added during the discussion session that every scientific message is not simple, every lesson is not easily digestible, and every result is not intuitive. Making complicated scientific results bite sized, palatable, and competitive with all other media risks undermining the work. Messages need to be propagated without undermining the integrity of science.
For communicating science to nonexperts, narratives can be appropriate and meaningful communication tools, said Michael F. Dahlstrom, assistant professor in the Greenlee School of Journalism and Communication at Iowa State University. However, narratives also raise ethical considerations about when and how to use them to communicate science.
A narrative is a causally linked temporal sequence of events involving specific, human-like characters, said Dahlstrom, adding “You might also call it telling a story.” Narratives are processed differently than are evidence-based arguments. The latter are context free. A fact can be removed from an argument and it still maintains its meaning. But a section cannot be removed from a narrative without losing its meaning and ruining the narrative.
Evidence-based communication begins with abstractions that can be applied to predict or explain specific situations. Narratives have the opposite direction of generalization, starting with specifics from which abstractions can be surmised. In fact, people will generalize from a narrative even when it is not representative of reality.
These two paths are not created equally, said Dahlstrom. Narratives are a critical way people make sense of the world, understand cause and effect, and interpret why people act the way they do. They are recalled twice as well and read twice as fast as evidence-based content. Calls to increase the teaching of evidence-based communication and logical reasoning are well placed, but they should not be interpreted to mean that narratives are an inferior way of thinking. On the contrary, narrative thinking may have given humans an evolutionary advantage by enabling them to figure out what others are thinking and might do.
Narratives in Science
The controversy over vaccines and autism provides a stark example of differences in the two ways of thinking, Dahlstrom noted. No evidence exists to link vaccines and autism, but narrative accounts of children demonstrating signs of autism a month after getting a vaccine can be very powerful. As the deputy director of the National Immunization Program once stated, “This is like nothing I’ve ever seen before. … It’s an era where it appears that science isn’t enough.” Rather than lacking trust in science, Dahlstrom noted much of the conflict is likely due to the narrative and evidence-based information being comprehended through different processing pathways.
Narratives also matter for science communication because many nonexperts get most of their science information from the media, and espe-
cially from television and the Internet. Most of this information takes the form of narrative-based stories, since narratives are a format that elicits attention among audiences. As Dahlstrom observed, the media try to focus on personal actions, fit events into a meaningful time frame, and personalize abstract concepts, which all lend themselves to narrative treatments.
Entertainment also uses narrative formats, and modern technologies have made entertainment ubiquitous. Someone may watch a half hour of news on television and then watch three hours of entertainment, with most of it in the form of narratives.
The Ethics of Using Narratives
Narratives are intrinsically persuasive, said Dahlstrom. They imply a normative assessment while neither stating nor defending their assumptions. They also reduce the formation of counterarguments by transporting a viewer, reader, or listener into the narrative. The more people are engaged, the more likely they are to accept what the narrative is telling them. Furthermore, fictional narratives result in similar levels of persuasion as do nonfictional narratives, which is one reason why the National Academies has established the Science and Entertainment Exchange to work with Hollywood film and television creators on fictional narratives.
These observations raise serious questions about the ethics of using narratives to communicate science. First, is the underlying purpose for using narrative improved comprehension or improved persuasion? Is the appropriate role of science communication to persuade an audience to accept views about science or to clarify understanding and engage a wider public in a more vigorous debate? These are completely different goals, Dahlstrom said, but a narrative can be used for either. A narrative can aim to persuade by emphasizing the preferred side of a science issue through characters that either agree with the preferred side or learn to do so through the narrative. Or a narrative can aim to increase comprehension by using events that explain all sides of a science issue and portray a character who is neutral to the issue or multiple characters who embody the different sides of an issue.
The second question concerns the appropriate levels of accuracy to maintain. Narratives can have multiple levels of accuracy. In some cases, accuracy may be relaxed for the larger purposes of communication. For example, a character’s motivations or actions, the settings, situations, events, procedures, and time frames may be more or less accurate and realistic and may all be used to communicate science. This happens even in science classrooms—for example, in discussions of frictionless surfaces, which do not exist in everyday life.
The representativeness of science narratives is a related factor. An audience will generalize from a narrative. Therefore, should the example chosen in a narrative be representative of a broader issue, or is it acceptable to use an outlier on which to base a narrative? The vaccine controversy is an example. Depending on whether the goal is increased understanding or persuasion, narratives may not be representative of reality.
The third ethical question is whether narratives should be used at all. They may violate expectations of how people think scientists should communicate. Science may be so linked with evidence-based communication that the use of narrative by a scientist may diminish credibility. Yet other stakeholders will likely be using narrative within the debate, Dahlstrom pointed out. Indeed, it may be unethical not to use narrative and surrender the benefits of a communication technique to the nonexpert side of a science topic.
Where Science Could Benefit from Narratives
Finally, Dahlstrom offered three questions regarding the future use of narratives in science communications:
• Do narratives help or hinder the desire to build trust between science and nonexperts?
• How can narratives meet the science communication needs of new media audiences?
• Can narratives help communicate science beyond human scale?
Human perceptual systems experience a very thin ribbon of reality. But to reason coherently about climate change, for example, people need to think in terms that extend far beyond a human lifetime. Narratives could help bring experiences that are outside of human scale within the realm of comprehension and consideration.
Countering Narratives with Narratives
Marty Kaplan, Norman Lear Professor of Entertainment, Media and Society at the University of Southern California’s Annenberg School for Communication and Journalism, pointed to another example of narrative-based promotion: direct-to-consumer drug marketing. Many companies are marketing drugs directly to consumers for things like depression, insomnia, and restless leg syndrome, accompanied by long lists of side effects. Many such advertisements are structured as narratives, where someone starts out troubled and ends up happy.
Scientific discourse competes with other narratives that are components of billion-dollar campaigns. Science needs to be involved in these campaigns, and it needs to counter narratives with narratives. As Kaplan said, “You know the expression ‘don’t bring a knife to a gun fight’? I submit, ‘Don’t bring a data set to a food fight.’”
The other discussant at the session, Melanie Green, assistant professor of psychology at the University of North Carolina, Chapel Hill, cited recent studies done in her laboratory on whether narratives or statistical information are more persuasive. Not surprisingly, people who use statistics are perceived as more competent, while people who use narratives are perceived as warmer. Other research suggests that the use of narratives can increase empathy for outgroups. To the extent that scientists are considered outgroups and want to be perceived as warmer, their use of narratives could increase the public’s receptiveness to scientific messages.
Narratives are analogous to science in that they deal with cause and effect, and one approach to writing a journal article, Green said, is to make it a good story. Finding the story in a data set can be useful in communicating science to a broader audience as well.
Finally, she noted that narratives are perceived in a social context. They are directed not just at individuals but at the social groups within which people live.
How do scientists actually communicate with each other, asked Kevin Dunbar, professor of human development and quantitative methodology at the University of Maryland in College Park. In the past they have used letters, journal articles, books, and presentations, while today they also have access to e-mail, Facebook, Twitter, and other new communication platforms. But all of these media have been built by humans as aids to the way the human brain works.
Analogies and Brain Activity
Dunbar and his colleagues have been studying communication in laboratory meetings in the United States and in Italy to learn exactly how scientists interact in those settings. They audiotaped and videotaped the meetings and interviewed the scientists before and after the meetings. For physician-scientists, they also compared interactions with patients to interactions with other scientists.
One prominent finding was that scientists relied frequently on analogies. They used local analogies within domains to fix problems, regional
analogies with nearby domains to generate hypotheses, and long-distance analogies across domains to conceive of explanations, with their goals dictating the kinds of analogies they use.
Dunbar and his colleagues also have been using brain scans to identify parts of the brain that are more active when making an analogy. For example, the farther an analogy is from the source concept, the more active is a particular part of the brain known as the front polar cortex.
Using Analogies Effectively
Dunbar also has been involved in brain-scanning experiments where subjects are given data that are consistent or inconsistent with a hypothesis. Data that are consistent with a hypothesis generate activity in particular parts of the brain, while data that are inconsistent do not.
The use of analogies makes a difference in laboratory interactions. Laboratories where members have similar backgrounds, such as a laboratory where everyone is a microbiologist, tend to have greater difficulty using analogies effectively. In contrast, laboratories where members have different backgrounds, such as a laboratory that combines physicists and chemists, can use analogies more readily to make discoveries.
Finally, Dunbar briefly mentioned that female scientists and male scientists do not differ in their use of analogical reasoning and social interactions. However, men were more likely to assume that they knew the cause of unexpected findings, whereas women were more likely to set out to determine the cause of such findings.
Using Stories to Communicate Science
Dunbar’s observations are a powerful argument for interdisciplinarity in science, said Green. They also shed light on how narratives can be used to reach particular types of audiences. Nonscientists often say that they cannot deal with math or that physics is too hard, but they do not say that they cannot deal with stories. Both analogies and narratives can make science more accessible to such individuals. At the University of North Carolina, for example, a program called “Scientists with Stories” is training scientists in storytelling techniques to help them better communicate their science while also helping them look for the stories in their own research.
Green also cited the importance of giving undergraduates research experiences so that they can learn that science is messy and hypotheses are not always confirmed. Even if they do not become scientists themselves,
they will know how the research process works, which could increase their trust in scientific findings.
Science as Narrative
Kaplan observed that Dunbar’s conclusions demonstrate that doing science is a narrative. Research has dead ends, surprises, mistakes, serendipity, and adventure. Even the choice of a problem to study involves ambition, competition, personalities, glory, and rewards.
Yet the drama of science is obscured in scientific papers, which are reverse engineered so that the outcome looks inevitable. Scientific papers are written from the perspective of “first-person invisible,” said Kaplan, with the process of science removed from the scientific results. Even though the process of science is a compelling story, scientists typically ignore that story in describing their work.
Narrative communications have a unique power to promote understanding, and that understanding can improve decision making, said Julie Downs, director of the Center for Risk Perception and Communication in the Department of Social and Decision Sciences at Carnegie Mellon University. Narratives can capture and hold people’s attention and provide the basis for a fuller understanding through coherent arguments, vivid imagery, and a foundation for new knowledge. They make people want to know what comes next, which means that people are more likely to get to the end of the message. People can acquire a general understanding from a narrative, even if they do not recall all the details. They can learn even when they do not realize that they are learning.
To translate science into narratives, theoretical models that can serve as guides are useful. In the health field, social cognition models of health are examples, though other models can also be used. These models do not provide specific content, but they broaden thinking and result in better communications than those created with no theoretical underpinning. To determine what content needs to be included, however, developers need to use a systematic investigation of what is known and understood by the target audience.
The Narrative Content
Narratives can take many different forms, some of which work better than others. The initial narrative is probably not going to be the best
communication. As a result, early versions of a narrative need iterative testing with members of the target audience. Are they understanding the narrative the way they should? Are audiences interpreting the word choices in a way consistent with the narrative’s objectives? To the extent that the narrative offers advice, how practical is that advice?
Pilot tests with a target audience need to encourage criticism so that the narrative can be refined and tested. The goal is a narrative that people understand in the proper way, that explains the science comprehensibly, and that urges action in the appropriate circumstances.
A Narrative Targeting Sexual Decisions
Downs used the example of a narrative that helps teens avoid pregnancy and sexually transmitted infection. The narrative was delivered through interactive video, which is an effective vehicle for audiences that may be skeptical and lack patience, which is the case for adolescents. Interactive video gives teens a feeling of agency and structure as they choose which way to go in the narrative. Teens also are used to nonlinear forms of media such as games or streaming videos.
The narrative was developed with expert input of what adolescents need to know to make good decisions about sexual behavior. Unlike much of the sexual education adolescents get, the narrative took a nonpersuasive approach. It sought to convey how infections are transmitted and how teens can reduce the chance of infection.
Teens are overwhelmed by what appears to be highly scripted behavior, said Downs. They adhere to behavioral scenarios that play out the same way every time, in the same way that people know what to do when they eat at a restaurant. Teenage girls describe going to a party, finding their way to a private room with a boy, and engaging in sexual activities. They do not see themselves as having much agency to act otherwise.
Teens also underappreciate relative risks and lack health knowledge. They have been taught in their sexual education classes that there is no such thing as safe sex, but they nevertheless will figure out ways to go right to the verge of what they have been taught not to do. They do not have a good understanding of what is high risk and what is low risk. They know about HIV infection, but have little understanding of how other sexually transmitted diseases are different and what implications that has for transmission or treatment.
The narrative builds on teens’ highly scripted behaviors to make them comfortable with the story. It has characters who follow scripted paths that pause several times with opportunities for decisions, at which point the narrative stops and the viewers are asked what they want to see the character do next. One option is to continue along the scripted path, but
other options would get the character off that path. The narrative also provides suggestions for how to take these alternate paths, some of which are cheeky and funny, others of which are direct or evasive. The videos then provide a 30-second cognitive rehearsal in which viewers can think about how to apply those suggestions in their own lives. “We can’t force them to think—if only we could—but we can at least force them to wait,” Downs said. “During this 30 seconds, we hope they give this some thought and apply it to their own lives.”
The videos also try to foster a better appreciation of relative risk through the metaphor of a risk scale that goes up and down. They point out that some behaviors are riskier than others and how to reduce the risks. They also explain reproductive physiology and attack misconceptions about, for example, how infections are transmitted.
A 6-month randomized controlled trial involving 300 subjects found that this approach resulted in decreased risky sexual behaviors and decreased sexually transmitted infections. A wider field trial was under way at the time of the colloquium that includes follow-ups and greater use of clinical outcomes and health records.
Taking Readers Out of a Narrative
A particularly intriguing aspect of this project, said Green, is its use of formative research to figure out what information people have and what information they need. That is a key step with these types of interventions that should be emphasized.
Green also played devil’s advocate with regard to the use of interactive videos. Despite the time and technology that goes into creating them, largely the same experience can come from reading a book. Communicators need to think about when interactive technologies are helpful and when it is better to stick with low-tech options. The nature of the audience is one factor in making this decision. Another is the psychological process a message is designed to evoke.
Narratives can transport a reader into another world, Green noted. Readers become immersed in the storyline and identify with the characters. But if readers have to stop and make a decision, they can be taken out of the narrative. The benefits of making them take responsibility for the future course of the story must be weighed against the potential disruption to the narrative experience.
The Role of Edutainment
Entertainment education, or edutainment, is a field that has been studied for 50 years, observed Kaplan. It has a highly developed theoreti-
cal base, a set of best practices, and techniques to evaluate its impact. It is well known for the impact it has had in such areas as combating adult illiteracy, domestic violence, and public health problems.
More than a decade ago, the Centers for Disease Control and Prevention (CDC) recognized that people pay attention to health messages in entertainment even if they know the entertainment is fiction. In 2001 it formed the Hollywood, Health, and Society program, which has been run by the Annenberg School and functions essentially as the CDC’s Hollywood office. The program has worked with hundreds of television shows to raise the profile of public health needs. For example, shortly before Atul Gawande’s book The Checklist Manifesto came out, the program brokered a connection with the show “E.R.” to have a life saved because a doctor was forced to use a checklist. The day after the show aired in New York, a conference of 150 surgeons watched the entire episode as a way to learn the value of checklists.
In “The Bold and the Beautiful,” a show watched by 500 million people worldwide every day, the Hollywood, Health, and Society program was involved in a storyline in which one of the main characters confessed to his fiancé that he was HIV positive. The day that happened, the STD/ HIV helpline spiked from 2,000 calls to 5,000 calls, a greater response than for every other public service announcement, campaign, and surgeon general’s announcement.
Note: This transcript of the final presentation on the second day of the colloquium has been edited for length.
Act 1: All the Science That Fits to Print
Ben dials on his phone, and Gabrielle answers her phone.
GAB – Hello?
BEN – Gabrielle Wong-Parodi?
GAB – Yes?
BEN – This is Ben Strauss from Climate Central. One of your colleagues at Carnegie Mellon recommended you as an expert on communicating risk.
GAB – I’m flattered. What are you looking for?
BEN – I’d like some help sharing results from a large study I’m leading on U.S. vulnerability to sea level rise and coastal flooding.
GAB – Tell me more.
BEN – We’re building an online tool to show our results. I think it’s very important that we share these results with the communities that could be affected most, and with leaders, and it has to be a powerful tool. I have the feeling that people really don’t get the danger that climate change poses, and that’s a problem I want to help tackle.
GAB – That sounds good. An online tool could be really valuable.
BEN – We’re generating a lot of data, and we don’t want to dumb it down.
GAB – Alright. …
BEN – The first thing is, we’re doing our analysis for every kind of place you can think of. We’re analyzing every coastal state, every coastal county, every coastal city, town, even zip codes. We’re looking at congressional districts, at state legislative districts, at federal and state agency districts, even at city council districts. We really want our work tailored for different audiences.
GAB – That’s a big positive. It’s well known that the more you can localize risk for specific audiences, the more you can command their attention.
BEN – We’re doing our analysis, too, for a huge number of potential impacts. We’re looking at housing, at population subgroups, infrastructure from power plants to airports to roads to rail; we’re looking at critical facilities like hospitals or fire stations; we’re looking at schools, churches, hazardous waste sites, military installations, parks, and more. Much, much more.
GAB – That’s a lot.
BEN – And then we also overlay our results against a spatial index of social vulnerability, divided into three categories based on a standard deviation method, so we can show how the physical exposure intersects with communities’ intrinsic response capacity. And we have results that assume levees, when present, are adequate in their protection, and a different set of results that assume that they are not adequate in their protection.
GAB – Alright. …
BEN – And, finally, most important of all, the time dimension. We made localized sea level projections at more than 50 water level stations, but also integrated them with local flood statistics to generate forecasts of flood risk, not just sea level. So we want to show sea level projections, annual flood risk projections, cumulative flood risk projections, plus how climate change multiplies risk, for each decade, and for 10 different water levels. We want to give users a choice of carbon emissions scenario, of sea level
model, and of what percentile estimate to view. How can we tweak our presentation so it hits in the gut? So people really get it.
GAB – I see. I would like to help.
BEN – I’m so glad I called. Talk again soon?
Act 2: Education or Manipulation?
Gabrielle calls Ben.
GAB – Hi, Ben. There’s something I want to talk about. You say you want people to get it. What does that mean?
BEN – That they understand the stakes. Feel them. And that the feeling fits the stakes and, ultimately, that action fits the stakes. I keep seeing surveys that show people ranking climate change low on their issue priority list. I know I’m a specialist, but that’s tragic to me. I want people to have the right level of concern.
GAB – Listen, I sympathize. And I happen to agree on the threat level. But did you hear yourself? How could you know what the “right” level of concern is for someone?
BEN – Maybe not personally. But at least professionally. Risk priorities should be in the same order as risk ranks.
GAB – But people are not always rational, at least in the way that you would like them to be, or they may have different values and priorities. You can’t just give them information and hope that it works. There’s so much more going on here.
BEN – That’s a big problem.
GAB – Albert Camus once said, “Fiction is the lie through which we tell the truth.” What if the only way to get people to feel the “right” level of concern—to really prioritize climate change according to its true risk—is to leave them with a false impression?
BEN – That’s a bigger problem. I need to stay true to the science. That’s my foundation. That’s who I am.
GAB – I needed to know that. I feel the same way. I’m not interested in spinning this.
BEN – So here’s the deal: we work together for compelling communication—as powerful as we can make it—that leaves the audience with a proper understanding of the science.
GAB – Deal.
Act 3: Optimist or Pessimist?
Gabrielle calls Ben.
GAB – Hi Ben. Ben, you really have to slim down how you show projections. This isn’t the control panel of the Starship Enterprise!
BEN – But I think it’s important to give people a range of values and some choice depending on how much risk they feel they can tolerate. Besides, the answers depend on future carbon emissions—on top of all the model uncertainties.
GAB – Providing some simple, limited choices makes sense. Just not the control panel of the Starship Enterprise. I’d suggest giving three or four choices—say, on a spectrum from optimistic to pessimistic.
BEN – Optimistic to pessimistic. That seems like it could capture a lot of things—uncertainty in the models, the level of emissions, and maybe even luck. And maybe it would give more of a personal connection.
GAB – That’s what I was thinking. But first I would like to test it.
BEN – Really? We’re just talking about a couple of words here. I like your intuition.
GAB – You would be surprised at the power a couple of words can have—and how wrong intuition can be.
Several months later.
BEN – Well?
GAB – Wow. Some surprises. If there’s one thing I’ve learned in my experience, it’s that data often trump intuition. “Pessimistic” made people think that the situation was very bad, that it was a worst-case scenario, as I expected. However, it also made them think that the situation was hopeless, that nothing can be done about sea level rise. Listen to what one subject said: “All of our coastal communities and development are doomed.” Then she added, “Seems as though fast rise can be dealt with, but ‘pessimistic’ makes me feel like nothing can be done.”
BEN – Really.
GAB – And “optimistic” seems to make people think sea level rise isn’t a problem at all. Listen to this: “That there’s hope, it’s okay that the sea level is rising, because it’s rising slowly and we won’t see any dramatic change soon. If we’re optimistic about slow rise, then we don’t really have to care.”
BEN – So, we get rid of our bright idea, and keep the terms simple: “fast” or “slow.”
GAB – That’s what the data say.
Act 4: Near or Far?
Gabrielle calls Ben.
GAB – So the results are in for the first big experiment using our simple research tool. It turns out that concern is highest for the persons to whom we showed the year 2050 projections, not the 2020 or 2100 projections.
BEN – A rather balanced outcome in light of the conflicting forces we imagined.
GAB – People tend to care about the near term: the right here and right now. The farther off the problem is, the less they worry—all else equal.
BEN – But the farther off into the future that we project sea level rise, the more dangerous it becomes. So how do those opposite trends play against each other? How steep or accelerated does the sea level rise curve have to be before it evokes concern for the more distant future?
GAB – Good point. I think 2050 may be our sweet spot because it’s far enough off to have a real sea level rise effect, but not much more than a 30-year mortgage away. It’s a stretch, but maybe people can imagine themselves in 2050, or somebody they know. But 2100 just seems too far away, no matter how awful the scenarios are.
BEN – Okay. So we established in this experiment that projections for 2050 evoke more concern. But is that concern appropriate? How well did subjects actually understand the risk, and did it vary by treatment?
GAB – People really understood the 2050 numbers well. Equally as well as the 2020 figures. I was honestly quite surprised; I’ve rarely seen such good comprehension on a survey. However, there’s a lot of confusion around the 2100 projections.
BEN – Well, that’s nice to hear about the 2050 numbers! I think it provides some powerful guidance for this project. I do have a nagging doubt, though. We tested subjects using projections and flood statistics for New York City. How robust are the results going to be for different risk profiles from different places? Does each place have its own sweet spot year?
GAB – That’s a really good point. I think I hear the sound of a new research proposal.
Act 5: Theory and Practice
Ben calls Gabrielle.
BEN – Gabrielle, we couldn’t do everything you recommended, or even that our work together suggested. We had only so much money, time, and flexibility. We did try.
GAB – I understand.
BEN – You advised me to simplify. The website and pages should be broken into bite-size pieces with a clear order. We landed on the idea of breaking the tool into four page types to handle four main functions, that we call WHERE, WHEN, WHAT, and COMPARE. That’s “mapping,” “projections,” “impacts analysis within communities,” and “threats compared among communities.”
GAB – I like it.
BEN – We broke the individual pages into sections, too. For example, with sea level rise and flooding projections in the San Francisco Bay Area, you can choose between the simple slow-through-fast sea-level rise scenarios we worked on labeling, how much carbon do you think we’re going to put in the air, how lucky do you think we’ll be, and how long do you think you’ll live?
GAB – I like what you did. A few simple choices up front, like we discussed, but an “Advanced” tab, too, that’s a bit harder to access.
BEN – Couldn’t help myself. Users who want can choose specific sea level models, emissions scenarios, and low-range to high-range results given those parameters.
GAB – That’s fair. Actually, it will be quite useful, I think, for some audiences. However, I’m a bit concerned that there will be too many options for some users and that it might be a bit confusing.
BEN – We did our best to explain, but I agree there’s a risk.
GAB – We talked about focusing on one year—2050—but also about how using a simple sentence, versus a chart, would improve most people’s understanding. As a scientist, it’s easy to forget sometimes how hard it is for people to understand numbers.
BEN – You know I’m not completely sold on 2050 yet; and honestly I didn’t really see how we could do it given the overall structure of the app. I take your points, though. We’re planning to use those findings in other contexts—like one-page fact sheets and press releases. Here it just seemed important to provide richer data. But if you hover your mouse
over one of the columns, you get a pop-up that explains the finding in a sentence or phrase.
GAB – That’s terrific! I like the simplicity of the description, too.
BEN – Well, I didn’t forget your lectures about writing for a junior high reading level. You chopped up too many of my sentences after your program evaluated them as written for philosophy majors.
GAB – Well, I hope people of all reading abilities can use this tool, and especially kids who are in junior high. They’ve got more at stake here than adults do.
BEN – That is too true. Look, I’m so grateful for your help, and can hardly believe that we’re almost there. We’re almost ready for our launch. Now comes the biggest test of all: how will people respond to the real thing, when it’s really about their backyards.
GAB – I can’t wait to see.
In the final session of The Science of Science Communications II colloquium’s second day, Dietram Scheufele, the John E. Ross Professor in Science Communication at the University of Wisconsin–Madison, identified four themes that struck him forcefully over the course of the day.
The first involves the role that scientists should play as arbiters of what is knowable. With controversies over vaccines, for example, scientists can determine the probabilities of certain things happening given particular levels of vaccination in the population. However, the policy implications of this information must be worked out through the democratic process, not in the scientific arena.
Second, the social and behavioral sciences have a fantastic new source of information in the data being generated by social networks. By making the invisible visible, these data provide scientists with information that they have never had before.
Third, collaborations involving business, scientists, and science communicators offer great potential, and not only in areas where science can help business sell more products. The field of science communication has much to learn from business that could be both unanticipated and extremely useful.
Finally, ethical issues play a surprisingly large role in science communication. What can be done is not always what should be done. Science communication needs to be held to a higher standard than most other forms of communication. That may put science at a political disadvantage, but failing to maintain high standards puts science at risk.