Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
4 STRATEGIES FOR THE FUTURE In this chapter we look ahead and beyond the more immediate challenges described in Chapter 3. Since its creation by Commissioner Hardy in October 1986, the Office of Strategic Planning (OSP) began work on what is now referred to as the Agency Strategic Plan (ASP). This document (U.S. Department of Health and Human Services, 1988) was developed to move the Social Security Administration (SSA) beyond its Systems Modernization Plan (SMP) by setting forth the agency vision and initiatives for providing services to its clients into the first decade of the next century. One of the tasks for the first phase of our study is to provide the SSA with our assessment of its ASP in light of the SMP,s legacy. We were asked to assess the ASP to advise whether it provides good guidance and how it might be improved in future versions. The following section has this objective. Our other objective for this chapter is to highlight several strategic directions that we believe are most promising for the agency's information systems. AGENCY STRATEGIC PLAN The ASP is a 48-page document that begins by stating the SSA's missionland its operating priorities. It describes the socioeconomic, demographic, economic, and technological trends and forces that affect the agency's clients and programs and provides a narrative view of the agency looking back from the year 2000. The ASP also lists 29 strategic recommendations that affect its programs, service delivery, technology, organizations, and human resources. It also provides a plan for implementing the strategic recommendations as a series of projects that can be completed by early 1991 and another set of interdependent projects that will proceed in four phases. In developing the ASP the SSA drew support from private industry, public interest groups, and its own people. The ASP is intended to be modified and expanded over time to reflect changing priorities and new initiatives. To implement the broad initiatives described in the ASP, the agency intends to produce supporting plans that include a 5-year tactical plan and a 2-year operational plan. Such planning is consistent with private industry, which also distinguishes among its strategic, 1 The agency mission, as stated in the ASP is: To administer equitably, effectively and efficiently a national program of social insurance as prescribed by legislation. 43
44 tactical, and operational plans. The strategic plans provide long-term direction ant! guidance to the organization. Its goals are based on visions toward which the organizations strive. Specific objectives to be accomplished are usually structured with time constraints in their organization's tactical plan. Detailed matters with budget figures for the next budget cycle are contained in the operational plan of the organization. Finally, in the planning process, major projects that require more than one budget cycle are laid out in specific project plans. The tactical plans typically have a time horizon of 5 years and set objectives accordingly. Operational plans typically set plans and budget projections baser! on specific projects and workloads for the present year and 1 year beyond. These tactical and operational plans were still being prepared during this period of our study, so we were not able to review them. The fact that these two plans were not available 17 months after issuance of the ASP suggests that the planning process is not functioning properly within the SSA. Our Assessment The ASP presents a broad and sweeping set of initiatives that go beyond just the agency information systems. These initiatives encompass the SSA's programs, client services, technology, and organization. Regarding its agency wide scope and approach, the ASP is well executed and effectively organized. The agency's mission and long-range strategic goals are succinctly stated. However, the ASP provides no real basis or justification for the goals it has chosen and in this regard, has a major deficiency. It also fails to present a broad vision of what the SSA of the 21st century will look like. Furthermore, the proposed initiatives and operating priorities are not explicitly related to the presented trends in demographics, economics, technology, and society. Future versions of the ASP should address these points. Strategic Goals The ASP lists six operating priorities that the agency's management selected to support the SSA's mission. Such operating priorities are effectively strategic goals that form an essential foundation upon which the subsequent initiatives proposed by the agency are based. These goals should be the basis for the tactical and operational plans that are being developed. Quoting from the ASP, the goals are listed as follows: 1. "Maintain the fiscal integrity of the Social Security trust fundsn; 2. "Improve public confidence in Social Security and how its programs are operates; 3. "Provide the best possible service to SSA's customersn; 4. NImprove management to facilitate greater effectiveness, efficiency, and accountabilityn; 5. "Use the best and most appropriate technology available to administer SSA programsn; 6. "Continue to insure that SSA can count on a properly skilled and highly motivated work force." Our greatest concern is with the third goal in that the degree of customer service is stated
45 without regard to cost or need. As was pointed out in Chapter 3, the SSA must clearly define and quantify the service levels' it intends to strive toward. Moreover, the SSA must carefully distinguish between the number and variety of services that it offers clients and the quality of each service. Services provided by the private sector tend to establish customer expectations. The SSA can provide customer service levels that closely follow those of the private service industries without having to exceed or even equal them. In the third goal listed above the words west possible. might be changed to Cast and accurate. in the next ASP. Considering the first goal, a board of trustees, which by law is composed of the Secretary of the Treasury as managing director, the Secretary of Labor, the Secretary of Health and Human Services, and two public members, is responsible for holding the trust funds and for making periodic' reports to Congress. The SSA, a constituent unit of HHS, has an important role to monitor the 'trust fund and call for changes when needed to assure fiscal soundness, but the agency ability to directly maintain its fiscal soundness is limited. The SSA can affect the trust fund's soundness by: (1) assuring that benefits are accurately paid, (2) controlling its own operating budget, and (3) influencing the Board of Trustees and Congress. In the first, although the SSA has a prime obligation to conserve the trust fund by accurately administering benefits, the fact is that even if the legislation were to grant benefits in excess of revenues, the SSA is bound by the law and its responsibility is to determine the recipient and amount of benefit accordingly. In the second, the agency's operating budget is such a small percentage (1.5%) of the trust fund that even with exquisite control of its budget, the SSA has minor leverage on the soundness of the trust fund. In the third, the SSA reports its views on the fiscal health of the trust fund to Congress, and through the Commissioner's position as Secretary of the Board of Trustees. To support its views, the SSA has an actuarial department that tracks and projects expenditures versus income for the fund. Regarding the fifth goal, we agree with the use of technology to administer the agency programs, but it must not become an end unto itself. We encourage the SSA to extend its use of technology as a means to better serve its customers in the most cost-effective manner, according to defined levels of service. We consider the second, fourth, and sixth goals to be entirely appropriate and within the agency's ability to accomplish. Planning Approach The SSA must clearly state a vision of what it is, where it is going, and how it plans to get there. Without such a vision of the future, planning for the application of new technology is speculative at best and without foundation. As a result, the technological choices that are made may solve immediate problems but may be inappropriate in the long run. For example, a strategic vision should address the number and location of SSA facilities and the functions performed at them, including the agency's district and regional offices, program service centers, and data processing centers. The ASP details the SSA's overall planning approach, focusing primarily on the needs of programmatic transactions. For-the planning function to be effective it will need such ongoing support and guidance from the commissioner that it becomes in effect a part of that office. The Office of Strategic Planning (OSP), established in 1986, is charged with both long- range planning and with coordinating the planning efforts of the various components within the agency. While this has raised strategic planning to the level of the commissioner's
46 executive staff, the challenge for the OSP is to provide leadership of the planning process agencywide and to support the other organizations in their planning efforts. This is a difficult challenge for a small and separate organization charged with a new function, but it becomes even more difficult in the long-established culture of the SSA. Missed Opportunities A typical strategic planning format used in private industry lists strengths, weaknesses, opportunities, and threats. For competitive organizations this format is appropriate, but for a federal agency it is more important to focus on opportunities first. While the initiatives in the ASP are broadly relevant to its goals, there are missed opportunities that should be addressed. Telecommunications Telecommunications is treated lightly in the ASP as a mechanism for service delivery or multi-Iocation voice and data connectivity. Yet in the SSA's report to the Senate Appropriations Committee (Social Security Administration, 1989), the fiscal year 1990 budget requested by the SSA for maintaining its current information systems includes S64.2 million for automated data processing (ADP) and $96 million for telecommunications. Since telecommunications, including the Data Communications Utility (DCU) and telephones, is budgeted at 150 percent of ADP costs, and hence dominates this part of the budget, we would have expected that the ASP would have included a strategic initiative to find ways to lower these costs without adversely affecting service or operations. Even without being able to precisely quantify what savings might be achievable, based on experiences of private industry, we believe that a multimillion dollar annual savings is possible by consolidating telecommunications procurement for both telecommunications and the DCU. This would best be accomplished by a central organization charged with procuring and managing the agency telecommunications infrastructure as is the common practice for major industrial corporations. Such central management in government is embodied in the concept of total information resources management. Central management is what the General Services Administration (GSA) accomplished in its Federal Telephone System (FTS) program known as FTS 2000. The GSA recently contracted with two major long-distance carriers to provide telecommunications services to federal agencies and the agencies are obliged to procure service through FTS 2000 unless good reasons to the contrary exist. The existence of FTS 2000 neither negates nor reduces the advisability for the SSA of centralizing agencywide responsibility for requirements, procurement, and overall management of telecommunications. Among the things we noticed in connection with SSA's handling of telecommunications: The SSA generally procures telephone and data communications as independent services, and even sometimes procures telecommunications on a programmatic basis, rather than an overall agency requirement. The SSA must address the combined telephone and data communications requirements and where possible consider them as a single item for procurement under FTS 2000 schedules. -
47 · For local telephone service there is no overall management of the interface between onsite equipment and the local telephone exchange. If this management existed it could help avoid buying more capacity than needed (overtrunking) and might also help avoid undertrunking, which impairs service. With constant monitoring of local usage statistics and proper management, a better balance of office demand and bunking facilities can be achieved and maintained. With over 1,500 separate facilities the possible savings to the SSA are large. We recommend that: The Social Security A.iministration consolidate voice and data requirements to take advantage of possible economies of scale in procurement and management of telecommunications facilities. Back Up and Recovery The SSA operates a highly centralized architecture for its data processing and storage. However, with one data center and a growing but unavoidable reliance on computer support for day-to-day operations, the agency needs to develop a strategy for dealing with the possibility of a sustained outage and needs to set goals to minimize the consequences of such an event. In this regard we note that had the agency chosen a format for its ASP that listed strengths, weaknesses, opportunities, and threats, this issue may have naturally appeared under threats. Chapter 3 addresses this issue. Automating Disability Processes The ASP should address increasing automation of its disability programs. For example, during our visit to the Office of Disability Operations (ODO), we found that this operation is based on moving, managing, and storing paper file folders. Modern technology clearly provides an opportunity for sharply improving the efficiency of administering the disability program. The disability program requires a disproportionate share of the agency's administrative staff in relation to the share of clients served. For example, only about 10 percent of the SSA's clients receive disability benefits, but this workload consumes about 32 percent of the administrative outlay from the trust fund. Later, in the section entitled Toward a Paperless Environment, we suggest technological approaches to address this function. AN ARCHITECTURE FOR INFORMATION Evolution of the Current Systems The current architecture of SSA's systems is the result of a gradual evolution toward on- line systems from a traditional batch processing environment. Present implementations and
48 choices in SSA's architecture appear to have been conservatively made and carried out by an incremental modification of previous systems, with new processing modes (e.g., on-line operations and telecommunications input from 800 numbers) being introduced in an adhoc manner in response to high-level strategic initiatives. The evolutionary heritage of past investments and decisions predisposes the agency toward centralized solutions for both its architecture and database structure. While such solutions have not been unreasonable given the past, the concomitant link to traditional ADP methodologies and attitudes has inhibited the exploration of more distributed hardware and software methodologies that would support the increasingly on-line environment for applications programs. Most limiting, however, has been the lack of agency initiative in planning for the eventual elimination of paper-based manual operations, which continue to consume a disproportionate fraction of costs, continue to impair productivity increases, and inhibit the smooth integration of newly legislated features. With the rapidly increasing technological advances in optical storage systems and image processing, it is important that the SSA plan for a transition to a paperless environment in the coming decade. A Framework for Future System Evolution The progress made under the SMP is significant. Nevertheless, the agency enters the early 1990s with the data processing methodologies of the 1970s and early 1980s. Before proceeding with specific information processing initiatives based on the goals of the ASP, the SSA should conduct a careful systems analysis of its existing functions, including a comparative study of how alternative information processing methods and degrees of automation will impact the performance of these functions for different levels of federal investment. Such an analysis is essential if the cycle of "catchup and crisis" behavior in information technologies is to be avoided in the future at the SSA. One possible way to view the factors pertinent to the analysis is to develop, for each of the SSA's major functions, a 3-dimensional matrix that displays: (1) component tasks, (2) processing requirements, and (3) possible levels of automation. For the latter there are clear groupings and a natural ranking: . . Paper-based, purely manual operation. · Paper-based, automated tracking/control, manual processing operation. Computer-based, batch processing operation. · Computer-based, on-line system input/batch output. · Computer-based, on-line input and output. Computer-based, on-line input and output and retrieval, fully integrated into other services. Every major function carried out by the SSA is actually a process made up of several subtasks, each of which may involve differing degrees of automation. As a result, the degree of automation possible for a function such as claims development or disability adjudication =%
49 might best be characterized by the mix or distribution of automation levels in these subtasks. For each of the subtasks it should be possible to also determine the personnel resources currently used, their degree of involvement, the data processing resources expended, and the average time of completion for the mix of actual jobs falling within the subtask for a standard time period. If statistics on-errors that lead to delay in job completion are also gathered, and the down-line effects of such delays are evaluated, it should be possible to assess costs and benefits for the current set of subtasks within each function. Besides the degree of automation involved in each subtask, it is important to characterize processing complexity in terms of access operations and processor utilization. Most of the on-line processes of the SSA currently involve fairly straightforward storage/retrieval processing and the generation of fixed summaries for individual SSA client files. At the district office level, on-line software is already being designed and introduced that performs some routine calculations that are appended to summary reports. However, the agency can and should go much further toward automating full processes, not just tasks, and reducing its reliance on paper. The SSA should press toward fully automating awhile persons processing from birth to death, including enrollment, accumulation of contributions, health care, retirement, and survivors benefits. A major shortcoming of current systems is that they do not permit concurrent access to several records from one functional program or multitasking of programs to dynamically generate problem-oriented files as needed. Such capability is essential for on-line handling of family groupings and claim/benefit assessments and calculations that, for example, might be determined by the nature and number of dependents involved in the grouping. In the future we expect that calculations, relationships among calculations and associated files, and operations involving multiple files all will become more complex and more often the rule rather than the exception. We anticipate that so-called rule-based expert systems will be used at local offices to handle some cases and some processing. At the regional and central levels processing of large numbers of records can involve more computation (e.g., statistical analyses and simulations). While such processing is entirely in the batch mode at present, we expect it to evolve into a mix of on-line and batch programs as more flexibility in operational planning is demanded in the future. In order to rationally plan for the smooth introduction of new technology, the SSA must augment its analysis of existing functions by a study of the effect of introducing advanced processing methods into not only subtasks but also the overall process. It should also consider whether modifications of existing subtasks and their sequencing or the introduction of entirely new and more efficient task decompositions are feasible, or even desirable or essential as higher degrees of automation are introduced into the function. Such analyses would exhibit and quantify the trade-offs among resources, costs, and benefits not only for the present level of automation but also for a variety of possible new process decompositions and associated levels of automation. Toward a Paperless Environment The process of producing, acquiring, tracking, retrieving, storing or filing, and managing paper documents consumes a major portion of the operating costs of any large paper-intensive operation. The SSA is no exception, and the introduction of on-line systems can do much to reduce the blizzard of paper. Even with the progress from the SMP, the bulk of the agency's processes still depend on paper file folders containing such items as original inquiries, internal
so and external correspondence, memos,and cuff notes from users. Such an observation reinforces our previously expressed view that the agency has automated tasks not processes; and the investment of the SMP has primarily enabled the agency to recover from a crisis with the actual increase in automation largely confined to the proliferation of computer terminals. On the other hand, there has been some progress in removing paper, and the SSA even refers to folderless systems. Nonetheless, file folders and their contents together with the file cabinets to store them still exist; they are just not accessed as frequently. Even with the progress from an all-paper to a paper-based operation supported by automation, the ODO, for example, has virtually no automation and depends on constantly moving huge volumes of paper file folders. The ODO is a challenging opportunity for automation because medical evidence and other information external to the SSA is Contextual (e.g., X-ray films). Such contextual information requires an imaging system to store it electronically, but integrated image and data storage and retrieval systems are relatively new to office and paper operations. Ongoing experiments by the United Services Automobile Association (USAA) and the Office of the Secretary of Defense provide some data. Excitement over optical technology is widespread; certainly it is a promising part of the long- term information architecture for the agency. For disability operations, image processing and optical storage and retrieval may be the key to any meaningful progress in providing on-line automated support to claims examiners and reviewers. The SSA is urged to undertake exploratory experiments and demonstration projects with such technology to examine relevance to ODO and to gain experience needed to guide the development of future systems. An issue of concern in both disability and nondisability operations is the legal status of image- versus original-copy retention. While such uncertain legal issues are a societal concern, the SSA could exercise leadership here in overcoming some of these obstacles in the next couple of years. The SSA should carry out an analysis of systems procedures and architectures needed to support paperless operations. Disability operations are a prime target for this technology. Toward an Architectural Evolution The SMP envisions a target architecture that represents the final stage of an advanced centralized processing system. It defined extensive on-line support to users, a well thought out and layered approach to software, and a large centralized architecture. In our view the agency should not stop at the stage defined in the SMP. Rather, it should consider the development of workstation applications to support the user and to share the workload with the centralized databases and processing facilities. We recommend that the SSA undertake the preparatory planning and analysis to select a proper scope of application of workstations that can be affordably implemented. The SSA is evolving from a batch environment where a user deals only with paper inputs and outputs to an on-line environment where many transactions and processes can be accomplished during a single on-line computer session. The evolution of today's system to an on-line environment will provide applications and database support to the users for each of the SSA's functions. However, if the user continues to operate a terminal supported completely by a central processor, the undesirable effects include: constraints on processing flexibility, constraints on sharing of the workload, continuing concentration of equipment at -
51 the National Computer Center (NCC), and excessive dependence on data communications. These can result in degrading overall agency performance. At a minimum, centralized support of remotely located computer terminals will be limited by capacity on the host system and by the communications network that links them. Worse, advanced features such as windowing techniques facilitating concurrent access to several client files at the terminal and user- oriented menus will be impossible or demand large increases in data communications support. We believe that workstation technology2 can clearly provide a broader range of user support and integration of function. It can provide local processing for applications and make possible advanced windowing and menu features. The workstation approach will focus central processing on database retrievals and updates that would make central processor workloads more predictable and stabilize the sizing of host processors and network facilities. We project that users at the program service centers and the district offices will require broader automation support, especially as the agency moves to a Whole persons level of automation. Transition to a workstation environment will allow significant sharing of processing with the central facilities. A careful requirements analysis will be needed because the workstation must span the full functional requirements of a user or set of users. On-line transactions with the NCC should emphasize data retrieval and file updating. Software development for workstation applications must be under central configuration control for both development and maintenance. Almost surely, local users will develop their own specialized applications, but the configuration management must assure that centrally provicled capabilities are not perturbed and that locally created features do not inadvertently subvert SSA policies and procedures. This should not be a problem as long as common procedures are reviewed and issued centrally. We recommend that: The Social Security Administration retain the present centralized database architecture but plan for the introduction of Intelligent workstations providing increased local support to the users of the system and embodying a common user interface for performing any agency function. Toward a Common User Interface ~ O As part of developing and introducing a workstation environment, the SSA should also develop a human-engineered user interface environment that spans the full range of applications, exploits contemporary approaches and features for such interfaces (e.g., pull- down menus, windowing), and interacts with the central processor transparently to the user. Such an interface, common to all users, can ease the introduction of workstation software and guide a user through complicated sequences of case actions. Such a development of a rich application set will take several years, but the basic features must be present initially so that subsequent extensions do not require relearning by users. Careful human factors analysis can identify the detailed features and suggest alternative 2 this term is intended to refer to high-performance microcomputers with autonomous capability. Workstation technology is currently available and rapidly becoming less expensive. Hi.
52 man-machine configurations for specific tasks and functions. An important payoff from a common user interface is the ease of moving personnel from job to job andisection to section, in turn minimizing some of the management headaches of personnel turnover. Toward an Integrated Architecture The integration of image processing, optical storage, and retrieval systems with workstation-based decision support systems promises to revolutionize office and paper operations in the next decade. From the standpoint of the facility manager, such a system can provide the receipt (date, time, source), routing (classification of action); tasking (assignment of responsibility and completion date); tracking (location and position in the queue); managing (past due, reassignment, repositioning in the queue), storage (archives); and retrieval of documents. It enables management to accurately measure workloads, reallocate work, or reallocate staff to complete work. Not only is management made more effective, but such systems save the additional costs of searching through piles of documents, manually filing and retrieving, and handling heterogeneous documents of different sizes. Such an integrated environment offers great promise for the SSA in all its operations, including disability workloads. The natural progress of automation at the SSA should extend from centralized batch to centralized on-line operations, to workstation environments, and to combined workstation and optical storage-and-retrieval systems. Such systems not only help management in assigning work, tracking and locating cases, and completing overdue actions, but they also help the individual user. Workstations can bring up (or Window inn) copies of all documents germane to the case at hand. The user can assemble on one screen all of the information necessary to complete a case action. The user can also manage his or her own queue of work, easily viewing assigned cases and their respective due dates and creating customized particularized management information such as cases worked on and those pending by type. The SSA should define a target architecture (initially at the top level) that uses available and near-term technology. Because of the cost and telecommunications demand for transmitting high-quality images, we project storage of documents and other images locally at the program service centers. While this requires a high-capacity local network to link workstations with optical storage devices, the long-haul data communications would not be seriously impacted. Management workstations would perform case control, workload management, and quality assurance activities. User workstations would access the NCC through a gateway off the local area network (LAN). Toward the Use of Expert Systems The introduction of expert-system technology promises greater productivity and standardization of processing for areas in which there exists well-specified decision criteria or at least sufficient expertise to identify and define "norms of practice." While there are various opportunities for automating processes via expert-system technology, disability claims processing and adjudication are the major prospective targets because of the present long delays in the process and the existence of standardized rules for large parts of it. The SSA has
53 already projected that automating the reconsideration and appeals process would reduce delays in disability processing from an average of 515 days to 299 days. The application of expert- system technology might help reduce the verification time and also improve the consistency of rulings on such claims, thereby reducing the number of cases overturned in the appeals process. In supporting an appeal one of the strongest points that a rejected claimant can make, and one of the most common arguments, is that a similar disability was allowed in another case. Such inconsistency in ruling on medical disabilities is understandable given that these determinations are made at many offices of state-operated facilities. By applying expert- system technology to make consistent rule-based decisions for disability verification, the verification time should be reduced and the percentage of appeals and decision reversals also should decrease. Laying the Groundwork While potential benefits from automating processes are clear, the need for expert-system versus conventional algorithmic approaches has not been established. Nor has the feasibility of using current expert-system technology in the more complex operations been established. The precision, unambiguity, and completeness with which decision criteria are known and the level of expertise in applying such criteria need to be examined as part of an initial exploratory prototype effort. In present expert systems (sometimes called knowledge-based systems) the state of the art is adequate to build systems for the SSA if the knowledge base (criteria and experience in using them) is adequate. There must also be a supplemental examination of present SSA functionality to identify potential applications of expert systems and, importantly, what is their likely interaction with conventional on-line and/or batch processes and with manual processes before choosing which decision areas are beneficial initial targets of opportunity. As part of this analysis, bottlenecks in the current manual paper-driven process must be-clearly identified and a human factors study performed to assess the feasibility of introducing different mixes of automation, including the expert-system components. Criteria for Applying Expert Systems In addition to disability processing, there may be other opportunities for introducing expert systems within SSA operations. These should be pursued only in conjunction with a clear plan that first makes full use of simpler automation methods for assisting on-line decision making at the district or regional levels. In general, the functionality of the SSA, in particular the most complex decision processes, must be inventoried and analyzed rigorously. One group of such categories is the following. As a reminder, Knowledge bases includes both the criteria for making a decision and the experience of applying those criteria. 1. Already existing knowledge bases of decision criteria that at most need some refinement. 2. Partially existing knowledge bases of decision criteria that need considerable knowledge engineering to convert to automatically usable decision rules.
54 Minimal formalized decision criteria but sufficient human expertise that can be codified into heuristics for the application and elaboration of the criteria. 4. No decision criteria but authoritative experts who claim to have consistently applicable decision heuristics. 5. No decision criteria but authoritative experts who apply a considerable degree of unformalized judgment in arriving at their decisions. 6. No decision criteria and only semisuthoritative experts who apply their judgment in arriving at conclusions. The first category above may well turn out to be implementable with conventional algorithmic techniques if the decision logic is sufficiently simple or restatable in a simple form. Even in such a circumstance which may not justify full implementation, a knowledge engineering approach can be useful to design a prototype by flexibly suggesting alternative ways of organizing and representing the knowledge. The second and third categories are amenable to application of current knowledge engineering techniques, with increasing degrees of investment in specialist time and knowledge engineering expertise over that needed for the first category. The fourth category will require truly intensive knowledge engineering and should not be undertaken at the present time unless considerable investment in automating the function is deemed desirable. The fifth and sixth categories involve ongoing research problems and should be set aside until new methodologies of knowledge representation and machine learning are developed. However, there might exist a different kind of opportunity for these two categories. The SSA might simply formalize (e.g., make explicit in a structured way) current rules of practice and decision processes that might then represent implicitly accepted methods of operation. In some cases such formalization may prove to be unproductive and may instead point to areas where legislation might need clarification or revision to change existing ambiguities or contradictions in the law. In other cases formalization may be advantageous to clarify administrative interpretations where these fall clearly under the executive jurisdiction. Suggested Steps We believe that the application of knowledge engineering, and in particular an expert- system approach to the business of the SSA, holds great promise; however, the agency must also recognize that this technology has limitations and should be approached in a carefully planned and deliberate manner. There has been overselling of the technology and overstatement of benefits; but on the other hand, there are many documented cases of significant improvements in productivity from relatively modest rule-based systems (Feigenbaum, et al., 1988~. ~~ Hence, we recommend that the SSA proceed in the following steps: 1. Assess the scope of problems first in the disability functions and subsequently in other areas to identify decision-making components amenable to prototype demonstration. -
55 2. Perform ~ cost-benefit analysis of the identified components and prioritize them as entrants in an overall Agency Strategic Plan. 3. Develop a prototype for test and evaluation. Goals of the effort must Include (a) a clear specification of benefits expected from different levels of decision support and degrees of automation, particularly as they relate to; (b) expectations of increased levels of service and support; and (c) outreach to the Increasing pool of clients who are Incapacitated, disabled, or have different cultural backgrounds. 4. Investigate the need for and benefit from formalizing current norms of practice in the Interpretation and application of possibly ambiguous statutory rules and criteria, with the view of either suggesting revisions in legislation or proceeding with the construction of formal lcsowledge bases when ambiguities can be resolved by administrative or executive action. Promising Advanced Information Technologies The SSA requires an ongoing program of technology assessment and concomitant planning to identify many opportunities that can arise from technology not now in use at SSA. For example, voice recognition, smart cards, optical storage cards, even the esoteric neural networks at some point in the future are all advancing sufficiently rapidity that the SSA shouIcl stay abreast of their capabilities and possible application to the agency's functions. For example, voice recognition systems could well increase productivity in the district offices but only after a full workstation capability has been installed to provide efficient and concurrent access to multiple records and programs. Voice recognition systems will eventually replace much of the computer system interaction now done through keyboard entry and also afford greater opportunities for the physically impaired in the workplace. The current technology is steadily improving the vocabulary of such systems and their ability to reliably adapt to different voices. Voice recognition systems trainable on a single user are within existing technology and should become economically feasible for incorporation into workstations by the time a workstation environment is widespread throughout the SSA. The smart card is a technology that embeds both microelectronic memory and processing in a plastic card. Typically the-size and appearance are similar to those of a credit card; the microchip memory can be read and altered electronically, and the internal processor can be used for many purposes. Smart card technology provides a convenient means for transporting large quantities of individualized client information. In the SSA context this could facilitate the processing of client queries and actions at geographically varied locations. The need for personal contact would be reduced, with resultant personnel savings. The savings would be especially significant if a client could access his or her information, of course with appropriate security and privacy provisions from home or public access points. By combining teleprocessing and smart card technologies, a variety of options can be envisioned to improve and customize individual access and service. Optical storage cards employ compact disk technology to store and retrieve information. They are relatively new but presently have the capacity of storing up to 16 million characters of information in a device the size of a credit card. An important difference between smart cards and optical storage cards is that the latter can readily store information such as data, image (photographs), signature, fingerprints, and X-ray images. At present, optical storage
56 card technology is a so-called write-once-read-many (WORM) device, so it is not currently intended for applications that require data to be readily alterable. It is an inexpensive technology to mass produce, and current estimates place the cost of a card at about 4 cents. In the long run the neural network might also have potential for the SSA. Since the neural net acquires its capability by Being taught~--in effect watching an individual(s) perform the process that it is intended to replicate--it is conceivable that decision processes for which an ill-defined knowledge base exists might be implemented with neural nets. Neural network technology is not now as well developed as expert systems, but its progress should be monitored. To summarize regarding advanced information technologies, we recommend that: The Social Security Administration study and inventory its technological base with a view toward establishing the relevance and candidacy of promising new technologies to its systems and needs, identify opportunities to exploit them to reduce costs and enhance services, and incorporate the results of such studies and examinations into revised editions of its Agency Strategic Plan. REFERENCES Feigenbaum, E. A., P. McCorduck, and H. P. Nii. 1988. The Rise of the Expert Company: How Visionary Companies Are Using Artificial Intelligence to Achieve Higher Productivity and Profits. New York: Times Books. Social Security Administration. February 1989. Report on Social Security Admi~zistration's Computer Modernization and Related Expenditures. Prepared for the Senate Appropriations Committee, Washington, D.C. U.S. Department of Health and Human Services, Social Security Administration Office of Strategic Planning. 1988. 2000: A Strategic Plan. Washington, D.C.: The Office of Strategic Planning, Baltimore, Maryland. -
