MODELS AND MODELING

Models are abstractions of reality, and modeling is the process of creating these abstractions of reality (Wallace 1994). Models take a variety of forms based upon their function, structure, and degree of quantification (Tersine and Grasso 1979). For example the functions of a chart of an organization is to describe and does not provide any predictions or recommendations; a sales forecast predicts the future based upon a particular business strategy; and a procedural manual for a manufacturing process is normative in that it provides advice on how to manage a process. The structure of a model can be symbolic (represented by equations), analog (using graphs to model physical networks), or iconic (physical representations such as scale models). Models are usually thought of as being quantitative, and able to be represented mathematically. However, qualitative models are far more common. For example, mental models play a very important role in the conceptualization of a situation (Crapo, Waisel, Wallace and Willemain 2000) and verbal and textual models are used in the process of communicating mental models. Because reality is near-infinitely complex, all data needs to be processed, which involves a movement from information to knowledge. Models are forms of codified knowledge.

Science can be seen as a model-building enterprise, because it attempts to produce abstractions of reality that help scientists understand the world (Little 1994). Technological advances in computing allow for the development of complex computer-based models in a wide range of fields. These models can be used to describe phenomena observed in the world as well as to provide structure to real or hypothetical experiences described or postulated by individuals or groups. Models play a very important role in formalizing and integrating theoretical principles from science that pertain to the phenomena being studied. For example, the computational models used for weather forecasting integrate scientific principles from a variety of the physical and natural sciences.

As the role of models within society increases, the significance of ethical issues related to the development and use of models also rises. Models are generally designed by experts who may hold privileged positions, yet model users and those affected by models may cover a wide demographic range. Thus, it is ethically imperative that researchers consider the relationships among the modeler(s), the model, the user(s), and those affected by the model.

Models may be developed for a range of purposes, in a variety of domains, including research, education, and applications. This entry begins with a brief overview of the ethics of modeling in each of these domains. The next focus is on ethical issues that span all three domains. Finally, the conclusion provides an assessment of the current status of the ethics of modeling.

Ethics of Modeling in Research

Models play an important role in scientific and engineering research. Scientific researchers seek to better understand the world, and models can serve as a way for them to create these understandings. Engineers try to improve the world by creating new technologies, and modeling allows them to explore their ideas in the abstract before moving on to the concrete. Computer aided design is one example. This technology allows an engineer to create a model design and view the resulting product in a three-dimensional graphical representation. This creative process can be repeated many times with various participants before the physical prototype is produced. In both science and engineering, models serve as tools for understanding the world and the ways in which people can improve that world.

One important ethical issue of modeling in research is the relationship between modeling and the norms of science. John Allison and colleagues (1994) argue that the fundamental ethic of science is an assumption of openness and access to data and methodology that fosters repeatability and verifiability. Yet, they point out that science increasingly relies on proprietary databases that do not allow others to repeat or verify the studies, such as economic analyses that use corporate financial data. They assert that models in this context may pose a danger to society, unless their data and methodology are kept open, as has been overwhelmingly the case over the long history of scholarly scientific research. Thus, it is important to consider not only the ends to which modeling is used in research but also the means through which it is used.

One way to ensure that models for research are used ethically is to develop a code of ethics for modeling within a particular domain of research. Saul I. Gass (1994) explores the codes of ethics for various research fields and organizations. He concludes that a uniform code of ethics should be developed so that researchers within a wide range of specialties can benefit from it.

Ethics of Modeling in Education

Another important use of modeling is for instruction. In education, models can be used to help students better understand a problem. Manipulation of the model—whether it is a formula, a plastic mock-up, or a computer simulation—helps students develop a better understanding of the problem at hand. Similarly, models can be useful in training, potentially allowing trainees to practice techniques and skills in a relatively risk-free environment.

Barbara Y. White and John R. Frederiksen (2000) argue that computer-based models are particularly important in education because they make scientific inquiry potentially accessible to all students. They assert that computer-based models can help students develop the conceptual models necessary for scientific inquiry. These tools allow students to experiment with models in order to better understand naturally occurring relationships captured by theories in physics or other academic subjects. White and Frederiksen further argue that students should be able to use computers not only to learn to apply models but also to create models and understand the principles behind modeling natural systems. Modeling in education can thus include both learning to build models and learning to use models.

Perhaps one of the most ethically intriguing applications of modeling is the use of virtual reality in education and training. Virtual reality models have been used to train and evaluate doctors, pilots, and other professionals. The goal of such models is to provide a safe environment that mirrors the work environment in potentially all ways except the consequences of the actions taken in the simulated environment. One issue requiring further study is the role that consequences play in affecting actions, and consequently, the potential utility of such environments. Another issue is that virtual reality environments may become so realistic that it becomes difficult or impossible to distinguish between the actual situation and the model of it. In such cases, transparency may be one way to avoid ethical dilemmas. Thorough documentation of the model, delineation of the assumptions the model makes about reality and values, and an explicit representation of the components of the model and how they are linked are all ways to help ensure the transparency of the model.

Ethics of Modeling in Applications

Modeling may also be used in a wide range of applications. Computational models have contributed to developments such as Dupont's discovery and use of ozone-friendly chemicals (Hoffman 1995), structures than can better withstand earthquakes (Booker 1994), and innovations in nanotechnology (Bozman 1993). Computational models are also increasingly being used for public policy-making (Kollman et al. 2003), and as a result they are receiving an increasing degree of attention in the popular press (see for example Ashley 2003). One major application of models is as aids for decision-making. Models used for decision-making may be either primarily descriptive or prescriptive—that is, they may attempt to portray reality as it is or reality as it should be. Neither of these tasks is as simple as it might seem. The design of both descriptive and prescriptive models is influenced by the perspectives of the participants, and thus it requires transparent communication and consensus between the builder(s) of the model and the user(s) of the model (Wallace 1994).

The relationship between the model builder(s) and model user(s) is inherently problematic. John D. C. Little (1994) describes six pitfalls for modelers to avoid:

1. The user already knows the answer and wants to use the model as a justification for it.
2. The user wants quick answers and does not give the modeler time to do a thorough study.
3. The user does not understand the basis for the modeler's results and thus is uncomfortable about using the model.
4. The user wants a defined, black-and-white outcome from the model.
5. The user is allowed to put her or his own personal judgments into the model.
6. The user does not realize that all models are incomplete.

Modelers must find ways to avoid these pitfalls that result from misinformed or misbehaving users.

Deborah G. Johnson and John M. Mulvey (1995) identify three types of relationships between modelers and users. First, they discuss a paternalistic relationship in whichthe modeleractsasanunquestionable expertwith total control of the relationship. Next, they explore a second way of understanding this relationship, the agency model, in which the user has the upper hand in the relationship, and the modeler is merely an implementer of the user's will. They reject both of these views as being unbalanced and failing to ensure that both sides strive to fulfill their roles. They conclude that the fiduciary model is the ideal model for the relationship between the modeler and the user, because under this model, the user and the modeler work together to construct the model and the user's expectations for the model.

Ethical Issues that Connect Modeling in Research, Education, and Applications

In each of these three domains of research, education, and applications, models can be used to either help or replace humans. Models used in research may either assist researchers or take over for them. Educators may either use models or be supplanted by them. Finally, in applications such as decision-making, models may either support human decision makers or automate their roles. Given this stark choice, it is important to consider the ethical implications of both models that help humans and models that replace humans.

Mulvey (1994) argues that models that are used to replace humans, which he refers to as "computerized decision procedures," are ethically problematic because they can easily be misused or abused. Intentional manipulation of a model may be used to serve the will of those that control it, who are often the elites within a society. Thus, models intended to replace humans may be used in antidemocratic and authoritarian ways.

Vincent P. Barabba (1994) points out, however, that models used by humans can also be misused and abused. A model can, for example, be oversold, so that limitations in the accuracy, precision, or scope of the model are underemphasized or completely ignored. In this way, models used by humans may also be used by elites to ensure that their will is achieved.

It is thus important to consider the power dimensions of models and modeling. As discussed above, there are a range of possible relationships between modelers and users, and the best type of relationship appears to be a fiduciary relationship whereby modelers and users each have both responsibilities and expectations as part of the modeling process. It is important that steps are taken to regulate this relationship, to avoid unethical behavior on either side of the transaction, and to ensure the best outcomes for both modelers and users, as well as for those affected by the model (Leet and Wallace 1994).

Models also present other ethical challenges. Models are designed to make reality more easily understandable, yet these same models may, intentionally or unintentionally, distort reality in important ways. Models may be used to make very value-laden decisions appear "scientific" and "objective." In building and using models, it is thus important to understand their limitations as well as the cultural specificity of the knowledge content and values that are explicitly and implicitly embedded in models (Leet and Wallace 1994).

Assessment of Ethics of Modeling

Richard O. Mason (1994) argues that modelers, as a part of their fiduciary relationship with users, have a professional responsibility for the models they build. To meet this professional responsibility, a modeler must fulfill two covenants: a covenant with reality and a covenant with values. The covenant with reality involves technical and social elements: The faithfulness of a model to reality often depends on highly technical decisions by the modeler, yet it is also a fundamental part of the relationship between the modeler and the user. According to the covenant with values, a modeler must understand and incorporate the user's values into the model in an effective way. These covenants are particularly important because a successful model may become a standard that affects a wide range of users and people affected by the model (see also Carrier and Wallace 1994).

In addition, it is important for the modeling process to be as transparent as possible. Because models always reflect the social and cultural context in which they are created, in both their knowledge content and values, it is most helpful if the model is open and honest about these influences. Models that contain assumptions should make these assumptions clear, rather than masking them as fact. Similarly, the extent to which a model is descriptive or prescriptive should be made immediately obvious to the user. Importantly, allowing the user to see clearly into the model is a way for the modeler to share control and responsibility with the user, allowing the user to make informed decisions based on all relevant data, rather than placing blind faith in a black box.

These three covenants—the covenant with reality, the covenant with values, and the covenant with transparency—can all help modelers and users communicate optimally so that they can mutually benefit from the process of modeling. All three covenants are important, because they make clear what users should be able to expect from designers, allowing designers and users to work as partners. Such cooperation ensures that modeling will be used for ethically responsible uses within the domains of science and technology.

WILLIAM A. WALLACE KENNETH R. FLEISCHMANN

SEE ALSO Georgia Basin Futures Project;Operations Research.

BIBLIOGRAPHY

Allison, John; Abraham Charnes; William W. Cooper; and Toshiyuki Sueyoshi. (1994). "Uses Of Modeling in Science and Society." In Ethics in Modeling, ed. William
A. Wallace. Tarrytown, NY: Pergamon. These authors address the question of the contribution and interpretation of models affects decisions.

Ashley, Steven. (2003). "Alloy By Design." Scientific American 289(1): 24. This brief article documents how the use of computational models led to the invention of a new class of titanium-based alloys, which are now used in medical implants, eyeglasses frames, and spacecraft parts.

Barabba, Vincent P. (1994). "The Role of Models in Managerial Decision Making—Never Say the Model Says." In Ethics in Modeling, ed. William A. Wallace. Tarrytown, NY: Pergamon. Barabba propounded Barabba's Law ("Never say the model says!") and discusses ways to limit model builders from making excessive claims for their models.

Booker, Ellis. (1994). "Working Toward a Quake-Proof Design." Computerworld 28(5): 86. Describes how computational modeling in an interdisciplinary environment is being applied to the problem of predicting the damage that can result from earthquakes.

Bozman, Jean S. (1993). "Tiny Technology: The Small World of Nanotechnology Opens Possibilities for Molecular Computing." Computerworld 27(31): 28. Touches on applications of computational modeling to nanotechnology.

Carrier, Harold D., and William A. Wallace. (1994). "An Epistemological View of Decision Aid Technology with Emphasis on Expert Systems." In Ethics in Modeling, ed. William A. Wallace. Tarrytown, NY: Pergamon. Discusses the philosophical foundations of statistics, operations research, and expert systems. Presents a framework to aid decision makers in choosing an appropriate decision technology for solving a particular problem.

Crapo, Andrew; Laurie B. Waisel; William A. Wallace; and Thomas R. Willemain. (2000). "Visualization and the Process of Modeling: A Cognitive-Theoretic View." In Proceedings KDD-2000: The Sixth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Boston, MA, ed. Raghu Ramakrishnan, Sal Stolfo, Roberto Bayardo, and Ismail Parsa. New York: Association of Computing Machinery. This paper provides a description of the process of modeling and, based upon theories of cognition, shows how visualization can assist in developing computational models for very large, high dimensional datasets.

Gass, Saul I. (1994). "Ethical Concerns and Ethical Answers." In Ethics in Modeling, ed. William A. Wallace. Tarrytown, NY: Pergamon. Offers guidelines for professional behavior by modelers.

Hoffman, Thomas (1995). "Making A Difference: Say Good-bye To Ozone-Wrecking Chemicals." Computerworld 29
(23): 105. Explains how computational modeling techniques have been used since the 1970s to explore the environmental impacts of ozone-depleting chloroflurocarbons.

Johnson, Deborah G., and John M. Mulvey. (1995). "Accountability and Computer Decision Systems." Communications of the ACM 38(12): 58–64. Focuses on the issue of accountability in the design of computer decision systems. Argues that it is important for the designers of computer decision systems to develop and adhere to a set of standards in order to increase public trust in these systems.

Kollman, Ken; John F. Miller; and Scott E. Page, eds. (2003). Computational Models in Political Economy. Cambridge, MA: MIT Press. Contains a variety of applications of computational modeling to the fields of political science and economics.

Leet, Edith H., and William A. Wallace. (1994). "Society's Role in the Ethics of Modeling." In Ethics in Modeling, ed. William A. Wallace. Tarrytown, NY: Pergamon. Discusses the proposition that values are inherent in any model, and offers suggestions on how to ensure users are cognizant of this fact.

Little, John D. C. (1994). "On Model Building." In Ethics in Modeling, ed. William A. Wallace. Tarrytown, NY: Pergamon. Reviews the process of modeling and notes several "pit falls" that can waylay the model builder.

Mason, Richard O. (1994). "Morality and Models." In Ethics in Modeling, ed. William A. Wallace. Tarrytown, NY: Pergamon. Presents three covenants that should guide the relationship between the model builder and the user.

Mulvey, John M. (1994). "Models in the Public Sector: Success, Failure, and Ethical Behavior." In Ethics in Modeling, ed. William A. Wallace. Tarrytown, NY: Pergamon. Focuses on the interpretation of models and the fact that lack of understanding of the techniques used in modeling hinders a user's ability to ascertain the values inherent in the models.

Tersine, Richard J., and Edward T. Grasso. (1979). "Models: A Structure for Managerial Decision Making." Industrial Management 21(2): 6–11. Presents a very comprehensive classification scheme for models.

Wallace, William A., ed. (1994). Ethics in Modeling. Tarrytown, NY: Pergamon.

White, Barbara Y., and John R. Frederiksen. (2000). "Technological Tools and Instructional Approaches for Making Scientific Inquiry Accessible to All." In Innovations in Science and Mathematics Education: Advanced Designs for Technologies of Learning, ed. Michael J. Jacobson and Robert B. Kozma. Mahwah, NJ: Erlbaum. The authors use the example of physics education to explore how modelng can shape student involvement in science.