As science attempts to answer bigger and bigger questions, it is more and more likely that the people participating in the effort together reside in different locations, institutions, and even countries. As noted in Chapter 1, scientific publications are increasingly written by teams and larger groups across institutional boundaries (Jones, Wuchty, and Uzzi, 2008). Geographic dispersion is one of the seven features that can create challenges for team science, particularly with communication and coordination. This chapter begins by delineating these challenges. We then describe, in turn, the findings of the literature on how these challenges are met by the individual members of the distributed team or larger group, the team or group leaders, and the organizations that wish to support distance collaborations.
Because many of the disadvantages that arise from being distant from one’s colleagues can be mitigated by various kinds of technologies, we next describe the suite of technologies available to support distance science. We then summarize how technology addresses some of the challenges of being geographically distributed. This chapter focuses on addressing a single feature of team science that creates challenges. Therefore, we do not include a separate discussion of the seven features that create challenges for team science as we do in Chapters 4 through 6. The chapter ends with conclusions and recommendations.
The chapter draws on many rich case studies of large groups and organizations1 composed of geographically distributed scientists and other pro-
1 As noted in Chapter 1, an organization typically incorporates a differentiated division of labor and an integrated structure to coordinate the work of the individuals and teams within it.
fessionals, which are supplemented by focused experiments and large-scale surveys and analyses of public records. For example, starting in the 1990s in the United States, the National Science Foundation has sponsored the development of a new organizational form for scientific collaboration called the Collaboratory (Wulf, 1993; Finholt and Olson, 1997)—a laboratory without walls. In Europe, this movement is called eScience or eResearch (Jankowski, 2009). To address science problems that are increasingly large and complex, collaboratories combine experts from multiple universities. Thus, they are typically geographically distributed, encountering all the issues outlined in this chapter in addition to those discussed earlier. The Science of Collaboratories Database (Olson and Olson, 2014) lists more than 717 such collaboratories, mainly in science but also in the humanities. Many of the entries include information about the topic, the participants, the shared instruments (such as the Large Hadron Collider) if any, funding, and the type of collaboratory, based on a proposed typology.
Challenges for geographically dispersed groups include members being blind and invisible to one another; time zone differences; differences across institutions, countries, and cultures; and uneven distribution of members across participating locations.
Being Blind and Invisible
People working with others at distant locations are both invisible to those colleagues (Bell and Kozlowski, 2002) and blind to their actions and situations. In addition, people working virtually with remote colleagues are often unaware of the detailed context of those colleagues’ work (Martins, Gilson, and Maynard, 2004). Research has shown that face-to-face communication is a valuable contributor to team performance (Pentland, 2012). Without explicit communication (Olson and Olson, 2000) or opportunities for periodic in-person visits, remote others do not know what individuals are working on, what their roadblocks and challenges are, and how they can help or be helped (Cramton, 2001). Technology solutions such as those outlined later in this chapter can help provide group members with the awareness they need to collaborate effectively, but group members must use these tools for this to happen. In other words, people need to take extra effort to report to remote others what they are working on, what the open issues are, and in general what the current context of work is, using e-mail, videoconferencing, teleconferences, or other electronic media.
There are additional issues of awareness not about the details of work but about the higher-level context of work. For example, a manager might unwittingly schedule a meeting during a remote location’s predicted blizzard, or, crossing country boundaries, during hours outside of their normal workweek (e.g., people in France typically work a 35-hour workweek, having Friday afternoon as part of the weekend). Conversations that include people at the same location may also include references to weather, politics, and sports familiar to the local participants, but not to those in remote locations (Haines, Olson, and Olson, 2013). Finally, people starting a virtual collaboration may have difficulty establishing a work norm, and individuals joining an existing virtual group may have difficulty learning and adhering to such a norm once it has been established.
Time Zone Differences
Scheduling meetings that include participants from around the world can be a challenge because of people working with collaborators in different time zones. Constraints on available meeting times can range from merely being an hour off to having no overlap in people’s working days (Kirkman and Mathieu, 2005). These constraints can lead to inconveniences to group members, such as the need to calculate and document accurate times among collaborators. Alternatively, some group members may have to make compromises to their own schedules, such as meeting early in the morning before their typical workday begins, during lunch, or late in the evening (Massey, Montoya-Weiss, and Hung, 2003; Cummings, Espinosa, and Pickering, 2009). Such compromises are more often made by the “minority” group member (the one individual on the other side of the globe) and can result in resentment or burn-out.
Differences Across Institutions
Science groups increasingly cross university boundaries. Academic institutions have different teaching schedules (some schools are on the quarter system, some semester, some intensive 8-week sessions). Different institutions also have different interpretations of rules about use of human subjects or about who owns intellectual property (Cummings and Kiesler, 2005, 2007). In addition, academic institutions use different technologies.
Differences Across Countries
Crossing country boundaries can create challenges regarding laws and expectations about intellectual property. In particular, regulations about the use of scientific specimens can differ, especially in human medicine. Laws
and expectations related to intellectual property differ not only in terms of ownership of discoveries, but also in terms of the use of ideas and writings of others, as expressed in different definitions of copyright and plagiarism (Snow et al., 1996). Expectations can also differ around protecting the privacy of human research subjects (e.g., by requiring individuals to sign informed consent forms), the use of data and software (with or without license), and how data are managed and shared.
Differences Across Cultures
Even more subtle than differences in laws and expectations about intellectual property are differences in unspoken norms of work, definitions of various terms, and work style expectations (Kirkman, Gibson, and Kim, 2012). For example, in the United States, organizational decisions are often made by a high-level steering group and then announced so that others will buy in. In Japan and India, the decision-making process is much more consultative, as decisions are worked out in small groups to gain buy-in before being announced more ceremonially to the whole organization (Gibson and Gibbs, 2006). Subtle factors about conversational style also can differ. For example, in some cultures, the pauses in conversation are long, allowing time to think and honor what was just being said; in other cultures (in particular the United States), conversations progress at a rapid pace and people may “step on” each other’s sentences and start to speak. A conversation including people from these two cultures can create impressions of disrespect on the one hand and assessment that the other has nothing to say on the other. Although beyond the scope of this report, there are many cultural differences when working across country boundaries, and these can have important effects on communication and ultimately effectiveness (Fussell and Setlock, 2012). Some very large, geographically distributed research organizations (e.g., CERN; see Box 6-1) provide support for these challenges, but other international groups are left to deal with these challenges on their own.
Uneven Distribution of Group Members Across Participating Locations
Often, members of geographically dispersed groups are not evenly distributed across all participating locations (O’Leary and Cummings, 2007). There is commonly a “headquarters” that involves the largest number of people, and satellites of one or two people included because of their special expertise. This is often referred to as the “hub and spoke” model. The culture and communication style of the headquarters typically dominate, and the group members at remote locations may experience lower status and
less power, while their needs and progress are invisible to others (Koehne, Shih, and Olson, 2012).
Power and attention are more evenly distributed if each location has a critical mass of people, although this presents its own challenges. As noted in Chapter 5, Polzer et al. (2006) found that having subgroups based on geography was associated with higher conflict and lower trust. In particular, conflict was highest and trust was lowest when there were two co-located subgroups (e.g., half of the group members were in one country and half in another). Similarly, O’Leary and Mortensen (2010) found that when there is a critical mass of participants at several locations, the individuals have a tendency to form “in-groups” and “out-groups,” with a tendency to disfavor and even disparage the out-groups.
As discussed in Chapter 4, individuals with social skills, such as those who score high on personality inventories as extroverts, are more likely to easily monitor and respond appropriately to actions and attitudes of others in their group or team (McCrae and Costa, 1999). Social skills are likely to be especially valuable in distributed groups, given that members need to communicate regularly and explicitly about the work being done.2 An additional individual characteristic that may be valuable is being trustworthy (Forsyth, 2010). Trust is an important binder of any group or team, and engendering trust is especially important when members have infrequent contact with each other and few opportunities to directly interact face-to-face (Jarvenpaa, Knoll, and Leidner, 1998).
As discussed in detail later in this chapter, working in a distributed group involves communication and coordination through collaboration technologies ranging from e-mail and audio/videoconferencing to more sophisticated systems for scheduling time and sharing documents or data. Thus another salient member characteristic is technological readiness—a disposition to learn new technologies and to access training to make the learning easy. Also required at the individual level is the openness to explore new ways of working, in which one explicitly communicates actions that normally require no special thought (Blackburn, Furst, and Rosen, 2003). In addition, the individual must be willing to commit the time needed to learn the new technologies, both to get started and then to share best practices as the technology is adapted to the work.
Because remote collaborators cannot see and interact with each other directly and may have to overcome divisive boundaries, they often must
2As discussed in Chapter 5, individuals can be trained to develop social and interpersonal skills.
learn new habits of working, many of them through technologies. In addition to good e-mail habits (e.g., acknowledging receipt of e-mail even though there is no time at the present to respond fully), people have to learn to explicitly make their actions available to others so they are aware of progress and obstacles (Cramton, 2001).
There is growing evidence that effective leadership can help science groups and teams meet the challenges of collaborating across long distances. For example, Hoch and Kozlowski (2014) conducted a study of 101 virtual teams and found that when teams were more virtual in nature, traditional, hierarchical leadership was not significantly related to team performance, whereas shared leadership (discussed in Chapter 6) was significantly related to performance. This result was expected because the lack of face-to-face contact and often asynchronous nature of electronic communication makes it more difficult for team leaders to directly motivate members and manage team dynamics. The authors also found that for these teams, structural supports were more related to team performance than hierarchical leadership. Structural supports provide stability and reduce ambiguity in ways that may compensate for the uncertainty that characterizes virtual environments. Such supports include providing fair and transparent rewards for virtual teamwork and maintaining ongoing, transparent communications while managing information flow. These and other leadership strategies that can help increase the effectiveness of virtual science teams are discussed below.
Leading Virtual Groups or Teams
One of the important leadership activities for distributed groups occurs in meetings. Meetings present a challenge because of the unreliability of audio/videoconferencing, and the lack of cues about who would like to speak next or people’s reactions to what is being said. The leader must explicitly solicit commentary and contributions from everyone, even polling individuals across all locations (Duarte and Snyder, 1999). This ensures not only that needed information and opinions are heard, but also that those at the smaller, distant locations feel respected for being asked. Also, when scheduling meetings among people who reside in disparate time zones, it is important that the leader fairly distribute the inconvenience of working outside of regular work hours to participate in the real-time meeting (Tang et al., 2011).
Also, the leader must be proactive in finding out what team or group members are doing (Duarte and Snyder, 1999). In a co-located setting, this
is done by informally walking the hallways. In a distributed science group or team, it requires regular contact with all members. Frequent contacts, by e-mail instant messaging, voice, or video, are critical to supplement more formal scientific or technical progress reports. This contact also helps members know that they are valued members of the collaboration.
Managing Group or Team Dynamics
The experience bases of individuals from different locations are likely to differ more greatly than the experience bases of individuals who are co-located. As discussed in Chapter 3, shared experience facilitates the development of two interpersonal processes that have been shown to enhance team performance—shared mental models (shared understanding of goals, tasks, and responsibilities) and transactive memory (knowledge of each team member’s unique expertise (Kozlowski and Ilgen, 2006). In addition, team members’ direct interactions shape team climate (shared understanding of strategic imperatives)—another process shown to improve team effectiveness. As such, virtual teams and groups are more likely to be successful if they engage in activities designed to overcome the lack of opportunities for shared experience, focusing, for example, on establishing common vocabularies and work style as explicit goals (Olson and Olson, 2014). This is especially important if the members come from different institutions and/or cultural backgrounds. Kick-off meetings are often used as a forum for members to explicitly assess habits and expectations, discuss differences, and agree on ways to resolve differences to increase chances for success (Duarte and Snyder, 1999).
Enhancing Readiness for Collaboration
To enhance readiness for collaboration, leaders can engage with members to foster intrinsic motivations, create extrinsic motivations, develop trust and respect, and thus improve group or team effectiveness. Individual members of a distributed group may have intrinsic (internal) motivation to work with the other members, either through personal ties or based on the realization that they need each other’s expertise in order to succeed. Both of these behaviors generate respect; when people feel they are respected, they are more likely to be motivated to contribute (Olson and Olson, 2014). If these conditions do not hold, then the leader may need to create extrinsic (external) motivators, including group rewards and individual incentives that reflect how well the person contributed to the group (discussed further in Chapter 8).
Activities designed to foster trust and team or group self-efficacy—two other team processes shown to enhance effectiveness in non-science teams—may bolster the chance of a science team or group’s success. First, because trust is slow to develop in a distributed group (with fewer occasions for people to learn how trustworthy others are and to become familiar with others’ personal lives), leaders could provide exercises or activities for developing trust. For example, virtual chat sessions, in which people are encouraged to talk about their non-work lives and share things about themselves that indicate vulnerability, have been shown to build trust (Zheng et al., 2002). Although such sessions can be valuable, the need to develop trust is one of the primary reasons that many teams conduct a face-to-face meeting of all participants at the outset of a project. Engaging the participants in team professional development activities can also build teamwork and trust (see Chapter 5).
The second, related interpersonal process that helps ensure success is team or group self-efficacy, an attitude of “we can do it” (Carroll, Rosson, and Zhou, 2005; see Chapter 3 for further discussion). This attitude encourages people to do extra work or find solutions when obstacles arise. Again, team-building exercises can help engender this attitude. As with trust, team self-efficacy enhances success in co-located as well as distributed teams, but when team members are distant, these processes are harder to establish and maintain.
Nature of the Work
When work is routine, such as on an auto assembly line, most people know what to do and what others are doing to coordinate their work. When work is complex, it is more challenging to keep track of what needs to be done and who is going to do which tasks. Collaborating at a distance is particularly difficult when the work is complex, as it is in team science (Olson and Olson, 2000). For example, in a study of 120 software and hardware development projects that were high in complexity, Cummings, Espinosa, and Pickering (2009) found that spatial boundaries (working across different cities) and temporal boundaries (working across time zones) were both associated with coordination delay. Coordination delay was defined in this study as the extent to which it took a long time to get a response from another member, member communication required frequent clarification, and members had to rework tasks.
One solution for managing complex work at a distance is to divide up tasks into modules so that most of the coordination and discussion happens among people who are co-located, essentially reducing the critical communication required across locations (Herbsleb and Grinter, 1999). Because of the stresses of distance to awareness, communication, and coordination, the
design of the work is critical (Malone and Crowston, 1994), and cognitive task analysis may aid in the distribution of work (see Chapter 4). If it is not possible to change the design of the work, then group or team members will be required to engage in extensive efforts to coordinate their research tasks.
Geographically distributed science teams and larger groups are typically composed of members from separate organizations (e.g., universities). The culture and incentive structures of these organizations influence the collaborative readiness of groups or teams that cross its boundaries. An organization’s culture sets the stage for the degree of competitiveness among, and status of, its members. The members within an organization work to act in ways that are aligned with reward structures. Misalignments, due to the incentive structure being individually focused versus team-focused or knowledge-driven versus product-driven, can have deleterious effects on its members’ ability to successfully engage in team science.
In academics, disciplines vary in their competitiveness. For example, some scientists conducting research on Acquired Immune Deficiency Syndrome (AIDS), such as geneticists, immunologists, and pharmacists, may be intensely competitive because of the large amount of money and prestige associated with finding a cure. In another example, scientists in the Bio-Defense Center, a consortium of organizations in the northeastern United States funded by the National Institute of Allergy and Infectious Diseases, did not initially share their data because of fear of being “scooped” by someone publishing findings from their data before the data originator could do so. Coordination of distributed work is always easier when a scientific discipline or community has a culture of sharing and cooperation (Knorr-Cetina, 1999; Shrum, Genuth and Chompalov, 2007; Bos, 2008).
In projects requiring individual scientists to submit data to a shared repository, reward structures (e.g., based on use of the data by others) may be needed to motivate people to share their data (Bos, Olson, and Zimmerman, 2008). GenBank, a genetic sequence database of the National Institutes of Health (NIH), requires genomic data to be entered into the database as a precondition for publishing. The Alliance for Cellular Signaling worked with Nature, a highly prestigious journal, to develop a new process to review and publish a database of “Molecule Pages” (Li et al., 2002). These datasets are the standard format for the output of hard work by the scientists, but differ from traditional publications. Nature editors would then certify this review process when young professors came up for tenure with these kinds of publications. In 2010, there were 606 Molecule
Pages published, 88 under review, and 203 under preparation (see further discussion of authorship, promotion, and tenure in the following chapter).
Competition can also play a role in scientific research. Not only is it a great motivator, but also it is the most immediate source of corroboration and error correction. Creating parallel teams is common in particle physics. For example, as detailed earlier in Box 6-1, two separate teams built and operated different detectors at the Large Hadron Collider in order to find and examine the Higgs particle. These large international teams worked independently and announced their results simultaneously, yielding two broadly consistent sets of results that have been accepted with high confidence.
The leader of a distributed science group or team is often affected by decisions made at the organizational level, such as the university. For example, incentive structures are often dictated by the organization, and the culture of collaboration and/or competition is often strongly influenced by the entire organization or even profession. The organization may dictate the design of the research project or designate how many people are located at each site, which can in turn affect how interdependent the tasks are, with the consequent stresses on communication and coordination. The funding agency or organization ultimately determines the project budget, which in turn dictates how much money is available for technical capabilities and support. Although the leader can argue for the importance of technology suites, support, and training to facilitate remote collaboration, the keeper of the funds often makes the final allocation.
When multiple organizations are involved, as is often the case in long-distance collaborations, there are additional issues to work out. Explicit efforts to align research goals across institutions may delegate the institution-specific goals to a secondary level. Legal and financial issues may have to be negotiated, for example, to reconcile varying approaches to allocation of project funds in different countries. In large academic research projects, there are issues related to who gets credit for the results, not just the publications, but at the organizational level, as well as who gets credit for the funding award and who owns the intellectual property.
Although many organizations seek to foster flexibility and creativity through a flatter organizational hierarchy, this approach works best for co-located teams, where it is easier to communicate and share context and tacit information. For large, distributed groups, work goes more smoothly with at least some authority and designated roles and responsibilities (Hinds and McGrath, 2006; Shrum, Genuth, and Chompalov, 2007). One recent study found that leadership that is shared and provides structural supports (e.g., providing fair and transparent rewards for virtual teamwork, managing information flow) improved effectiveness in distributed teams (Hoch and Kozlowski, 2014).
Organizations and group leaders may benefit from the use of an online assessment tool called the Collaboration Success Wizard, see http://hana.ics.uci.edu/wizard/ [May 2015]. This tool asks the participants in a particular team science project to answer approximately 50 questions about the nature of the work, the motivations, the common ground, the management, and the technology needs/uses in the project. The respondent can ask for immediate feedback on where the team or group is strong, where vulnerabilities might lie, and, importantly, what to do about them. Following completion of the surveys, project leaders can obtain a summary report, again showing strengths, vulnerabilities, and what to do about them, because there are occasions when different individuals or subgroups may have different views about their work.
In this section, we first review the kinds of technologies that have been used to support distributed work, with different kinds of work benefiting from different constellations of technologies. The committee’s framework follows closely that of Sarma, Redmiles, and van der Hoek (2010), categorizing technologies as communication tools, coordination tools, and information repositories, adding significant aspects of the computational environment (see Box 7-1). Although we refer to specific technologies, the point is not to recommend a specific current technology, because it will quickly be replaced with newer versions. We rather wish to emphasize the types of technology that are useful and why. We then present an analytic scheme to guide people in choosing the right constellation of technologies for their work.
Types of Collaboration Technologies
E-mail and Texting E-mail is ubiquitous, and many experts have characterized it as the first successful collaboration technology (Sproull and Kiesler, 1991; Satzinger and Olfman, 1992; Grudin, 1994; Whittaker, Bellotti, and Moody, 2005). One of the cornerstones of its success is that today it is independent of the device or application used to send and receive it, and, with attachments, it is a way to share almost anything the recipient can read. As happens with other technologies, people also use it for managing time, reminding them of things to do, and keeping track of steps in a workflow (Mackay, 1989; Carley and Wendt, 1991; Whittaker and Sidner, 1996; Whittaker, Bellotti, and Moody, 2005).
Classification of Technologies to Support Distance Work
E-mail and texting
Voice and videoconferencing
Chat rooms, forums, blogs, and wikis
Large visual displays
Workflow and resource scheduling
Laboratory notebook (online)
Large-scale computational resources
SOURCE: Olson and Olson (2014). Reprinted with permission.
Instant Messaging (IM), sharing primarily simple text messages with another person or even a group, has made significant inroads into organizations. In some cases, it has replaced the use of e-mail, phone, and even face-to-face communication (Muller et al., 2003; Cameron and Webster, 2005). There is evidence that it is sometimes used for complex work discussions, not just simple back and forth about mundane issues (Isaacs et al., 2002). It is also used effectively for quick questions, scheduling, organizing social interactions, and keeping in touch with others (Nardi, Whittaker, and Bradner, 2000).
Except for e-mail attachments (which can include elaborate drawings, figures, and videoclips), the technologies listed above are text-based, even
in the abbreviated world of texting. Text remains an impoverished medium compared to the tones and facial/body expressions possible in face-to-face communication.
Conferencing Tools: Voice and Video There are a myriad of opportunities to communicate beyond text in today’s world, and many are used heavily. The telephone trumps text in being able to convey tone and to have immediacy of response. However, delays caused by technical interruptions of voice and video transmission are highly disruptive to conversational flow because of the importance of pauses in turn-taking in a conversation (Börner et al., 2010).
Many people have telephones from which they can teleconference, at least on a small scale. Organizations often provide services for larger-scale audio “bridges” for conference calls. Key to the smooth execution of these calls is whether the phones have “full-duplex” or “half-duplex” transmissions. Half-duplex lines are capable of transmitting only one direction at a time. Natural conversations often include “backchannels”—the “uh huh,” “hmms,” and other comments that convey whether the recipient is agreeing, understanding, or not; when using a half-duplex line, these responses are silenced. As a consequence, often the speaker will talk longer than necessary, not sure if the recipient has understood (Doherty-Sneedon et al., 1997). Additionally, conversational turn-taking is often signaled by an utterance from the one who wants to take the turn while the current speaker is speaking (Gibson and Gibbs, 2006). These are entirely cut out in a half-duplex line, creating awkward competitions for who will speak next.
Although tone of voice can add meaning to the words said, facial expressions and body language add another layer. In large meetings, video helps convey who is present without an explicit roll call, and by eye contact and expression, conveys who is paying attention. One can see not only the people but also the situation or context they are witnessing.
The richness of voice and video, however, can create barriers to people who are from different cultures. As noted earlier, the expected pause structures in conversation are different in the Western and Eastern cultures, often creating miscues. Because Westerners are used to a shorter pause structure than Easterners, they will dominate the conversation (Hinnant et al., 2012). Similarly, when video shows facial expressions and eye contact information, because those modes of expression are interpreted differently in different cultures, people again may make wrong attributions of interest and consent.
For greatest effectiveness, a video connection should be arranged to mimic a sense of physical presence. Eye contact and gaze awareness are key linguistic and social mediators of communication (Kendon, 1967; Argyle and Cook, 1976). In video, as in real life, people tend to focus on the face of the person with whom they are talking and attempt to make eye contact
by looking at the eyes of the person. Unfortunately, to appear to make eye contact over video requires a person to look not at the projected eyes of the remote person but at the camera. Therefore, to convey eye contact, extra effort needs to be expended to move the video of the remote person as close to the camera as possible. Without this careful adjustment, meeting participants will appear as if they are glancing sideways or at the top of other participants’ heads, both of which can be interpreted as disinterest (Grayson and Monk, 2003).
Conversations are often accompanied by gestures referring to an object, a document, data, or a visual image. Today, sophisticated tools, such as GoToMeeting, Google Hangout, and Skype screen-sharing allow a participant to share his or her computer desktop or a particular window with others, allowing them to control what they are looking at and the ability to focus attention by using the mouse/pointer.
Blogs, Forums, and Wikis Longer conversations from larger numbers of people are usually accomplished through chat rooms, blogs, forums, and wikis. Chats are nearly real-time, whereas blogs, forums, and wikis have a longer time between contributions. When used for distributed science, all are typically restricted to a designated work group rather than being public.
The large groups of space physicists participating in the Upper Atmospheric Research Collaboratory and Space Physics and Aeronomy Research Collaboratory used chats extensively to converse during their “campaigns,” periods when the sun’s activity impacted the upper atmosphere. The automatically recorded chats allowed people to “read in” to the conversation (scrolling back and reading what had been happening), helping them “catch up” although their time zone differences prevented them from participating in “real time.” The conversations were comparable to those held face-to-face (McDaniel, Olson, and Magee, 1996).
Wikis similarly are free-for-all conversations, but are even less structred in formatting. Forums are typically set up for discussion threads, whereas wikis can take any form whatsoever. The large groups of scientists participating in the Biomedical Informatics Research Collaboratory used wikis extensively to share test protocols, tips, frequently asked questions, announcements of the availability of new software tools, and articles of interest (Olson et al., 2008).
Virtual Worlds Virtual worlds are graphical, 3-D representations of physical spaces and have drawn considerable attention from both industry and academia (Bainbridge, 2007). They allow a person to experience a realistic environment, usually through an avatar. Avatars can explore a space, manipulate objects, and, when networked together, interact with other people’s avatars. The Meta-Institute for Computational Astrophysics is a collabora-
tory based exclusively in virtual worlds. The institute provides professional seminars, popular lectures, and other public outreach events in the game Second Life3 (Djorgovski et al., 2010).
Such simulations of real worlds have been in common use for training in the military for a long time (Johnson and Valente, 2009). Although multiplayer games such as World of Warcraft4 also allow for a wide range of playful interactions, Brown and Thomas (2006) speculated that real leadership skills might be learned in a game such as this because it involves extensive quests with a substantial numbers of players.
A class of technologies exists to support collaborators in finding a time to work synchronously, and a second set of technologies supports coordination during their time together.
Shared Calendars Although the original introduction of group calendars was met with resistance, many organizations have seen value in their use (Grudin, 1994; Grudin and Palen, 1995). Calendars support the coordination of meetings, finding a time when the important participants are available.
Calendars are also used as a tool to display and/or read availability. When colleagues do not respond to requests in their usual timely way, one can view their calendars to discover whether they are out of town or in a meeting. The information also allows for planning when to contact a person (e.g., an “ambush” after an in-person meeting in order to get a signature). Shared calendars can be particularly valuable for geographically dispersed colleagues who are in different time zones, reminding people of when the workdays overlap and where they do not.
Awareness Tools Today, awareness information is conveyed in the status indicators of IM systems. With IM, the user has control over what status indicator to convey to others, but the feature comes at the cost of remembering to set it and actually setting it. The cost of receiving the status setting, however, is very low. Many IM clients list the person’s chosen colleagues who agree to be monitored, and their status is typically listed in iconic form on the edge of the screen.
IM indicates the user’s current state, from which others can infer whether she or he can be interrupted, but not specifically what they are
doing. In the domain of software engineering, a key form of advancement in science, where coordination of detailed efforts is of primary importance but the work nearly invisible, developers have created and widely adopted various system to “check out and check in” portions of the code they are working on. For example, Assembla5 is a collection of tools to track open issues and who is working on them, plus a code repository where code is assigned to a person to work on, during which time others are locked from editing. These kinds of coordination tools are powerful, but not widely adapted to domains other than software engineering.
A more general system that notes what people have been or are working on in a shared document appears in Google Docs. The names of others who are currently editing the document are shown at the top of the document, and their cursors with their names in a flag are shown where they are working now. In addition, the list of past revisions and an indication of who did what (with authors’ contributions highlighted in different colors) show what has been changed. These various symbols and colors provide awareness of others’ efforts on a common document, useful if more than one person is working on the document either at the same time or asynchronously.
Meeting Support Coordination support for meetings, whether they are face-to-face or remote, can be formal and informal. During the 1990s, developers and users tested Group Decision Support Systems, in which participants were led by a meeting facilitator through a number of computer-based activities such as to generate ideas, evaluate them in a variety of ways, do stakeholder analysis, and prioritize alternatives (Nunamaker et al., 1991, 1996/1997). But these systems fell into disuse because of their management overhead and cost.
Informal meeting support tools typically take the form of a simple projected interactive medium, such as a Word outline or a Google Doc. The outline lists the agenda items at the highest level in the outline; during the meeting, a scribe takes notes that everyone can view and implicitly vet. As agenda items are completed, the outline format allows the item to be collapsed, implicitly giving a visual sense of progress. Those applications that allow multiple people to author the shared document, such as Google Apps, are even more powerful in these settings. When there is a single scribe, that person typically is so busy that he or she is barred from contributing to the conversation. When there are multiple authors “live,” while one scribe talks, others can take over seamlessly to enter notes on what they are saying. Additionally, these note-taking tools have been used very effectively in teams that include people for whom English is not their
native language. The real-time visible note-taking is akin to “closed captioning” of the meeting.
Workflow and Resource Scheduling Routine tasks that require input or approval from a number of people benefit from a structured digital workflow system. A number of efficient online systems handle this type of flow. For example, a very successful workflow system supports the National Science Foundation grant submission, review, discussion, and decision-making process, notifying the appropriate players in the process at the appropriate time, giving them the tools and information they need, recording their actions, and sending the process on to the next in line. Although the rigidity of these systems can sometimes prevent their adoption, a number of such systems have succeeded (Grinter, 2000).
In some research endeavors, especially in the natural sciences where the expense of a large piece of equipment necessitates researchers sharing it, systems have been put in place to schedule time on the equipment. For example, time allocation for use of telescopes is managed with software systems created with the joint goals of being fair to those requesting time and maximizing the use of the equipment. Bidding mechanisms have been explored to optimize various aspects of the complicated allocation problem (Takeuchi et al., 2010). Various kinds of auctions have been tested to both create an equitable distribution of time and to prevent people from “gaming” the system (Chen and Sonmez, 2006).
Whether a science team or larger group is co-located or distributed, it often needs to organize and manage shared information. The model of informally collaborating by sending people edited documents as attachments is common but fraught with challenges. Issues of version control and meshing of changes emerge. A better solution is to have a place where the single document resides as a shared file, with all the authors having access. Microsoft, for example, offers Sharepoint, an integrated set of tools selected for file sharing. It includes collections of websites and collaboration and information management tools (including tools for tagging documents for permissions and types and automatic content sorting). It also allows search through all the contents. To date, however, the system has not been widely adopted by research universities, which are using a range of different collaboration tools.
Another example of a system for shared editing and file management, but with a more fluid form, is Google Apps and Google Drive. The applications within Google Apps (documents, presentations, spreadsheets, forms, and drawings) each can be shared with others or placed into a folder,
which also can be shared via Google Drive. This set of features gives the users flexibility, but without vetted “best practices,” many are not using the applications effectively. The variety of different “cloud” technologies for document sharing is confusing to users (Voida, Olson, and Olson, 2013). As individual scientists and research institutions adopt various tools, the lack of interoperability sometimes forces scientists to revert to the “lowest common denominator” of sending documents as e-mail attachments (see Box 7-2).
Scientists who share data rather than documents face an additional set of challenges related to data quality, data-sharing, and database management (Borgman, 2015). If data are being collected by a science team or large group then the members have to agree, at the outset, what constitutes good quality data. Many large science groups have goals that include sharing data across sites. For example, in the early development of the Biomedi-
User-Centered Design for Collaboration Technologies
Technology intended to support virtual collaboration sometimes does not support it and even poses a barrier to collaboration (Crowston, 2013). Unless the technology is chosen or designed to both fit the users’ needs and be easy to learn and use, it will not support the collaboration. A collaboration tool that requires extensive training, is difficult to use, does not fit collaborative activities, or does not work well with other technology is likely to interfere with collaboration and may eventually be abandoned. User-centered design can help technology adapt to the users, not vice versa.
Developing technology to fit the users’ needs requires careful analysis of the users’ tasks, infrastructure, culture, and overall work context. Beyer and Holtzblat (1998) outline the steps in such an analysis to ensure that the technology has the right functionality. They consider
- communication flow,
- order in which steps occur in the work,
- artifacts produced and used in the work,
- culture, including power and influence, and
- physical layout.
Once these are made explicit, the people making the decision about what suite of technology to use (whether it be purchased or created) can brainstorm and then design the final solution.
When then designing the user interface to the various technologies in the suite, Norman (2013) proposes six principles:
cal Informatics Research Network (BIRN), the participants believed that progress on understanding schizophrenia would benefit from having a larger sample size of magnetic resonance imaging (MRI) images of patients, both with and without schizophrenia, doing various cognitive tasks while being imaged. A great deal of effort was spent in ensuring that the tasks that the patients performed were standardized and that the various imaging machines were calibrated. In other large groups of scientists, great care was given to developing a shared ontology of medical terms so that patient data could be aggregated from different locations and from different medical specialties, each of which had its own vocabularies (Olson et al., 2008).
In some domains of science, the laboratory notebook is a key tool for recording and vetting information. The researcher uses the notebook to keep a personal record of daily activities, such as tests run, information gathered, and observations. It is important to sign and date each entry
- Consistency: Similar technologies should work in similar ways; users should not have to learn new procedures for each new piece of software.
- Visibility: Controls should be clearly marked and not hidden from user view.
- Affordance: Form and other visible attributes of the technology should intuitively guide function, (e.g., clickable elements of the interface should be highlighted).
- Mapping: There should be a clear and evident relationship between controls and their effects (e.g., as when volume on a slider bar increases if the bar is moved up or to the right).
- Feedback: Effects should follow actions immediately and obviously.
- Constraints: User options should be restricted when unavailable or inappropriate (e.g., grayed out when not allowed).
Whittaker (2013) notes that successful use of technology often relies on following best practice, but it is unclear how users are expected to learn best practice. A single system seldom does everything a group or team needs: one is for workflow and scheduling, whereas another is for storing and sharing information. Interoperability problems abound, as when data-sharing tools to do similar work operate in different ways. Users are not equally familiar with the components that make up systems, and frustration can cause people to fall back to lowest-common-denominator technologies such as e-mail or spreadsheets.
Research is needed to improve the design of collaborative technology for team science. Such design would benefit from the philosophy outlined in the Human Systems Integration approach that puts the human at the center (National Research Council, 2007a).
to record important discoveries, often feeding into patent applications. Noting the value of being able to store and share these notebooks, some large scientific collaborations have developed electronic notebooks. The Electronic Laboratory Notebook (ELN) developed at the Pacific Northwest National Laboratory (PNNL) was so well designed that it was used heavily throughout the labs and adopted by other collaboratories even in different domains (Myers, 2008).
Aspects of the Computational Infrastructure
The System Architecture Many large groups of scientists have no choice as to how to architect their systems. The large-scale computation technology is either local or hosted on a private grid of secure machines, and, at NSF-funded centers, the data, often large, are stored on their own massive servers. At a more fundamental level, only a few large research projects can afford to create their own data storing and sharing systems; many scientists still rely on Microsoft Excel software.
Those scientists who have no need for storing or computing with massive data have a choice of whether to purchase applications for installation on their machines or to opt for computing and storage “in the cloud.” If choosing to work in the cloud, then connectivity is important if collaborative access in real-time is required. Many cloud-based applications offer some level of off-line activity, although the availability of up-to-date version control is lost. A more serious concern for some is security. There is resistance to cloud computing among clinicians, military contractors, police and fire departments, certain government agencies, and others who are sensitive to information loss.
One interesting consequence of these different architectures is that each architectural choice creates its own behavioral consequences. When the applications and documents are on private machines, the mode of collaboration is hand-off, serial revision: Documents are revised with “tracking changes” on and sent to the author-editor, who in turn can choose to accept each change or not. The power resides in whoever the collective has made editor. In contrast, where the document and application resides “in the cloud,” there is an implied place where those designated as editors can go to make changes. In this model, each edit appears as if accepted; the document is changed. Others can view the revision history and undo the changes, but at least at present, a reversion to an earlier version undoes all changes, not just one at a time. Neither model in its current form is ideal.
These are two entirely different modes of collaborating in terms of workflow. Often collaborators tacitly make the decision about who has the power to make changes, who can merely comment, and who has the final say in accepting the changes proposed. The existence of these two models
presents additional challenges to the users who are involved in collaborations of both kinds. They have to remember where something is stored, how to find it, and who has the power to decide on edits in each case, a situation referred to as “thunder in the cloud” (Voida, Olson, and Olson, 2013).
The Network Underlying all collaboration technologies is the network. Simply put, the bandwidth has to be sufficient for the kind of work to be done. Most of the developed world has adequate bandwidth for ordinary tasks, including video. Specialized needs that require large amounts of bandwidth will require specialized network infrastructure. Many large scientific projects have had to build high-performance networks to handle the volume of data that comes from their instruments as well as specialized computing to garner enough resources to do the computation on that mass of data. For example, the ATLAS detector at CERN produces 23 petabytes6 of raw data per second. This enormous data flow is reduced by a series of software routines that lead to storing about 100 megabytes of data per second, which yields about a petabyte of data each year. A special infrastructure is required to manage data flows of this size.
Large-Scale Computational Resources In many areas of endeavor, such as advanced scientific research or data mining in business, large-scale computational resources are needed. Certain high-end centers, such as the National Center for Atmospheric Research, have traditionally developed their advanced computational resources in-house. But organizations such as NSF, realizing that there is a need for advanced computing in many areas they serve, have supported the building of infrastructures to support advanced computation. The historically important supercomputer centers are one manifestation. A particularly noteworthy example of advanced infrastructure to support such needs is the Grid, a sophisticated computational infrastructure that is widely used (Foster and Kesselman, 2004). A more recent example is the NanoHub,7 a special computational infrastructure for nanoscience and nanotechnology.
Human Computation There is also a tradition of using human capabilities aggregated over large numbers to achieve important computational outcomes, often called “crowdsourcing.” Although there are examples of this as early as the 1700s, the phenomenon has experienced a recent renaissance under other rubrics (Howe, 2008; Doan, Ramakrishnan, and Halevy, 2011), such as collective intelligence (Malone, Laubacher, and Dellarocas, 2010),
6A petabyte is 1015 bytes. As reference, 103 = kilobyte, 106 = megabyte, 109 = gigabyte, 1012 = terabyte.
the wisdom of crowds (Surowiecki, 2005), and citizen science (Bonney et al., 2009; Hand, 2010). The core idea is that in many domains, gathering together the small inputs of a large number of individuals (“micro tasks”) can lead to results that can be as high in quality as judgments by experts and done in a fraction of the time.
In sum, science groups or teams typically need technologies to support communication and the sharing of the objects around which conversations take place. Technologies are needed to coordinate the conversations, both to find times to converse and to coordinate around the objects. The objects, information, and/or data, need to be collected to exacting standards, managed, and made accessible. Underlying it all is the architecture and networking, and large-scale computation occasionally supplemented by aggregated human computation. Effectiveness happens when the tools needed are available and used appropriately by the group or team members.
Selecting a Constellation of Technologies to Meet User Needs
New technologies often fail to live up to their promise, and it is not always clear what underlies the success of certain technologies, though these factors seem to include active leadership, deployment strategies, and how a particular tool fits in an overall assemblage of tools (Whittaker, 2013). Therefore, which technologies are chosen for a particular science team or group, and how these technologies are managed, can have an impact on the success of the collaboration. In selecting a constellation of technologies for a virtual team or group, it is important to consider the following factors (Olson and Olson, 2014):
- speed of response, impacting conversation and immediacy of data understanding;
- size of the message/data or how much computation is required, impacting required computation and networking;
- security, impacting choices about architecture;
- privacy, again, impacting choices about architecture;
- accessibility, impacting who can easily get access;
- richness of what is transmitted, impacting conversation and data understanding;
- ease of use, impacting adoption;
- context information, impacting coordination across sites;
- cost, impacting what can be accomplished; and
- compatibility with other things used, impacting adoption.
Choosing the appropriate suite of technologies to support a science team or group is not easy. The features of each technology drives how it will be used and often dictates social configurations of use. Although we have not provided a decision tree to guide selection of the “right” set of technologies, we have provided a listing of classes of collaboration technologies and the key features of these technologies that should be carefully considered in the choice of one’s particular use. It is important to consider all facets of collaboration at a distance: communication, coordination, information repositories, and computational infrastructure.
We next consider some examples of how technology and particular social practices can address each of the challenges we have identified to remote collaboration
Being Blind and Invisible
Videoconferencing and awareness tools can be used to increase visibility of participants as well as display who is working on what. Because it is important to communicate explicitly about the nature of work to be done as well as to share contextual information surrounding the work, videoconferencing can provide a feeling of presence for remote members and permit gestures, linguistic cues, and other ways to enhance communication among virtual team members. Awareness tools that permit the use of status indicators (such as IM) or color coding of document changes (such as Google Docs) can also be beneficial. Of course, they are only effective if the people involved invoke them, keeping the video on, setting their status markers to indicate their availability, and using the issue tracking systems.
Time Zone Differences
Whether a few hours or a full working day apart, scheduling meetings and coordinating work across time zones can be a challenge. A shared calendar, when used by all members of a virtual team, can greatly reduce the time spent scheduling teleconferences and work-related conversations. The calendar can signal when team members are working on parts of the task in addition to highlighting when they have free time available for casual conversations about the work. The shared calendar also serves as a form of documentation of the times members regularly meet, and especially for those across time zones, can reinforce norms around regular meetings.
Scheduling meetings including people whose workdays do not overlap can still create imbalance in who has to be inconvenienced. This is a social issue that has to be worked out with the participants and their management.
Differences Across Institutions
Typically, group members at different institutions are subject to different protocols, database access, and calendars. Workflow and resource scheduling that incorporates different institutional priorities, policies, and procedures can make coordination needs of participants salient. When group members from two different institutions do not share the same academic calendar, have different protocols for Institutional Review Board approval, or have different levels of access to online databases, coordination challenges can arise. Through a workflow and resource scheduling system that documents which group members are responsible for which tasks, who has access to particular sources of information, and what approvals are required and when, the institutional differences can be made explicit and accomodated. Systems that allow members and leaders to keep track of activities across institutions and provide notifications when action is required should facilitate coordination for multi-institution science groups.
Coordinating work around all of these differences requires explicit discussion. Successful distance collaborations often begin with a “communication covenant” that outlines the differences across institutions and the procedures the participants have agreed upon to coordinate.
Differences Across Countries
One of the best tools for determining how laws, rules, and policies vary between countries is a broadly accessible information repository such as a wiki. Groups that use such information repositories can document and track changes in regulation and intellectual property laws as they are occurring. Because all members have access to the latest information posted on the wiki, and can add, modify, or delete as necessary, the task of keeping national information up to date is shared across group members.
Differences Across Cultures
Today, English is the lingua franca of international scientific collaboration involving U.S. institutions. Much confusion and misunderstanding can follow from an understandable failure to appreciate linguistic nuances especially when spoken by remote members of large groups. Written communication, through e-mail, texting, and “chat rooms,’’ allows people to write out what they are thinking, and, furthermore, allows other members
to read (and re-read) the message to process what it means. Members from different cultures might find text-based communication more effective than real-time, voice-based communication.
In addition, suites of tools such as GlobeSmart have been designed to educate people about their collaborators’ cultures and behaviors and to find a middle ground.8 For example, if one is from a culture where the manager typically makes important decisions, she or he will be surprised when a collaborator hesitates in agreeing with the manager because everyone is consulted before a decision is made in the other culture.
Uneven Distribution of Members Across Participating Locations
Skillful use of meeting support technology can facilitate and broaden participation in decision making (e.g., by distributing a dynamic agenda), build procedural fairness (e.g., through electronic voting) across sites, and reduce power differences. When a majority of members are at the headquarters with a few other members scattered across different sites, it is easy for the remote member to feel isolated and in the minority. Meeting support technology, such as having a common Word document with an agenda that gets annotated as the meeting progresses, can ensure that members from all locations get heard (and recorded). A PowerPoint slide that outlines the procedures for voting on a decision, or even indicates who is going to lead the meeting (which can switch each time), can put the virtual group or team on the same page. The use of WebEx and other tools for running distributed meetings that integrate voice, documents, slides, and other materials facilitate the inclusion of members from different sites, big and small. These tools exist; it takes a manager open to contributions from all participants to use the tools effectively.
Large groups of scientists, as well as smaller science teams, are often geographically dispersed, requiring scientists to rely on information technology and other cyber infrastructure to communicate with distant teammates. Addressing the special challenges facing such teams requires effective leadership and technology.
CONCLUSION. Research on geographically dispersed teams and larger groups of scientists and other professionals has found that communicating progress, obstacles, and open issues and developing trust are more challenging relative to face-to-face teams and larger groups. These
limitations of virtual collaboration may not be obvious to members and leaders of the team or group.
RECOMMENDATION 4: Leaders of geographically dispersed science teams and larger groups should provide activities shown by research to help all participants develop shared knowledge (e.g., a common vocabulary and work style). These activities should include team professional development opportunities that promote knowledge sharing (see Recommendation #2 earlier). Leaders should also consider the feasibility of assigning some tasks to semi-independent units at each location to reduce the burden of constant electronic communication.
CONCLUSION. Technology for virtual collaboration often is designed without a true understanding of users’ needs and limitations, and even when a suite of appropriate technologies is available, users often do not recognize and use its full capabilities. These related problems may thus impede such collaboration.
RECOMMENDATION 5: When selecting technologies to support virtual science teams or larger groups, leaders should carefully evaluate the needs of the project, and the ability of the individual participants to embrace new technologies. Organizations should promote human-centered collaboration technologies, provide technical staff, and encourage use of the technologies by providing ongoing training and technology support.