Springer 2005
Topoi (2005) 24: 221227 DOI 10.1007/s11245-005-5057-1
Playing with Ethics: Games, Norms and Moral Freedom*
ABSTRACT: Morality is serious yet it needs to be reconciled with the free play of alternatives that characterizes rational and ethical agency. Beginning with a sketch of the seriousness of morality modeled as a constraint, this paper introduces a technical conception of play as degrees of freedom. We consider two ways to apply game theory to ethics, rationalist and evolutionary game theory, contrasting the way they model moral constraint. Freedom in the rationalist account is problematic, subverting willful commitment. In the evolutionary account, freedom is also problematic, leading to an infinity of possible social norms with dubious normative force. While these two approaches complement each other, we argue that the evolutionary approach is superior on both theoretical and practical grounds.
1. Introduction ‘‘play … 12. To move freely within limits, as parts of a machine’’ (Webster’s New World Dictionary)
Ethics is serious yet involves agents who play with alternatives. Rational agents consider a range of alternatives, moral or not, and ethical agents must be able to choose among a range of moral options. So good models of moral and ethical mechanisms should account for these tensions. This paper addresses this problem in four steps. In section 1, we characterize seriousness as a constraint and freedom as play. In section 2, we introduce and contrast two ways of applying game theory to ethics, which we label rationalist and evolutionary game theories, in terms of how they account for freedom and moral seriousness. In section 3, we consider some practical differences between the two approaches. We conclude that both theoretical and practical reasons suggest evolutionary game theory is the more promising method for our research program to model ethical mechanisms.
Peter Danielson
1.1. Moral seriousness Morality is serious; we sanction moral deviants. Indeed, on our account informed by game theory, norms are only stable if reciprocal, where reciprocity includes the willingness to sanction the non-compliant (Danielson, 2002). Nonetheless, to count as rational and ethical, the agents to whom morality applies must be free to consider moral and non-moral alternatives. The practice of philosophical ethics reflects this freedom to consider alternatives. The seminar room is not a church and beginning students of philosophical ethics are often surprised, if not shocked, at the irreverent attitude of some of the philosophers they read and discuss there. Perhaps the clash between serious subject matter and playful methods explains why ethics is the branch of philosophy with an esoteric tradition. Plato, Sidgwick (1893), Strauss (1952) and Parfit (1984) all argue that openly reflecting on morality is disruptive of established social order, so philosophical ethics must be hidden from the hoi polloi, or in Parfit’s case, even politicians.1 Consequently, in their technical literature, if only rarely before wider publics, philosophical ethicists will (for example) speculate whether people have any more right to life than lab animals.2 They are engaged in what Schrage (2000) calls ‘‘serious play’’. It touches core values guarded by powerful psychosocial forces like fear and envy as well as morality and other institutions.3 We propose modeling seriousness as a feature of the constraint morality imposes. Morally serious agents not only refrain from doing what is wrong, they refrain from taking wrong options lightly, and, for seriously wrong actions, perhaps even considering them as options. This suggests modeling seriousness as a measure of constraint. One may violate a not very serious rule for one’s own reasons, but more
222
PETER DANIELSON
serious rules require consulting with others, justifying one’s action, through to avoiding the action and, in the extreme case, even to avoiding considering noncompliant alternatives. 1.2. Play: Options and choice Given the seriousness of the subject matter, why call the methods of ethics ‘‘playful’’? We deploy the word ‘‘play’’ with an analytic focus: ‘‘play’’ denotes degrees of freedom. Play here derives from the mechanical sense (quoted from Webster above) projected into the space of ideas. For example, adding a ‘‘universal’’ joint will allow a rigid shaft to be ‘‘bent’’ by adding two new degrees of freedom of movement. Extending the metaphor, opening up new options extends freedom of thought and perhaps action. We conjecture that play in this sense models part of the problem of moral freedom. A precondition of rationality is the ability to consider the play of alternative options. (This broad claim extends from accounts of rational decision that start with the set of options, to accounts of learning as involving generation and testing of alternatives). Yet morality demands that some attractive alternatives are not chosen, others never chosen, and still others perhaps are not even considered. They are ‘‘off the table’’, else their consideration be indicative of a morally corrupt mind (Hampshire, 1978). It is important to note that we need not go to extremes to invoke the constraint of seriousness on intellectual freedom. Teaching in an applied ethics program, we have had graduate students express concern when pressed to attempt a utilitarian justification of alternatives to their professional ethics. Such speculation is risky and they were concerned, appropriately, with their reputation and status as trustworthy moral agents.
2. Rationalist and evolutionary game theories This section sketches two strands of recent game theory applied to ethics in a way relevant to our problem. We contrast what we call the rationalist game theory (RGT) of Gauthier (1986) and McClennen (1990) with the evolutionary game theory (EGT) of Binmore (1994, 1998a), Gintis (2000a), Skyrms (1996, 2003), Sugden (1986), and Young (1998). RGT and EGT have much in common (hence
the ‘‘GT’’ in their acronyms) that should be welcomed by ethicists: a focus on real social interaction, the motivation to be moral, information transfer, and, overarching all these, contact with the progressive sciences of economics, evolutionary biology, and computational social science. Indeed, the close relation between evolutionary theory and game theory generally provides a welcome generalization of game theoretic tools useful for ethics (Danielson, 2004). None the less, rationalist and evolutionary game theory differ in crucial respects. They self-identify as the game-theoretical projections of Kantian and Humean approaches to ethics, respectively.4 Simplifying greatly, RGT sees ethics in terms of a simple issue turning on a mistake about rationality; EGT sees ethics problems as symptomatic of the limits of reason to capture social complexity.5 We consider how each addresses our problem of free play of alternatives vs. moral constraint. For Gauthier and RGT generally, ethics is conceived in Kantian terms, contrasted with rational self-interest, and takes the prisoner’s dilemma (PD, see Figure 1) as its core problem. In contrast, Binmore and Skyrms shift the methodological focus to Hume and the canonical problem to one with multiple equilibria. 2.1. RGT: Self constraint We begin with RGT and sketch how it models our problem of moral freedom and then consider the solution offered by RGT’s leading author, David Gauthier. The well-known prisoner’s dilemma is a promising model of the problem of moral freedom. It is a situation where having additional alternatives makes players worse off, so they have an interest in having less alternatives, that is, in mutual constraint on their freedom to choose. To see how this works, consider the PD in Figure 1. Here both players can each receive the (Pareto) optimal payoff of 2 by choosing C, but, tempted by the greater non-coop-
C
D
C
2,2
0,3
D
3,0
1,1
Figure 1. The Prisoner’s Dilemma.
PLAYING WITH ETHICS
erative payoff of 3, they each end up getting the lessthan-optimal payoff of 1 by choosing D. So it appears that if they did not have the tempting but illusory non-cooperative diagonal alternative, they would have been better off. Gauthier’s proposed solution, ‘‘morals by agreement’’, recommends that they agree to renounce the freedom to choose these options. To make it easy to sketch Gauthier’s solution, consider an even simpler case: an extended PD game where Row moves first and Column follows. Since Row knows that Column has a better option choosing D, Row should also choose D. However, were Column able to commit herself to choosing C (if Row does), and were known to have done so, Row would prefer to choose C. This solution exemplifies Gauthier’s principle of constrained maximization. Column is constrained (conditionally) to cooperate. (In this asymmetrical extended game, Row need not be constrained.) Yet constrained maximizers remain maximizers. Indeed, too much constraint say a willingness to choose C unconditionally does not lead to the cooperative outcome (Danielson, 1992). This appeal to maximization is the reason that this approach is seen as a competing variation of game theory. Straightforward maximizers will not be able to solve single play PDs by themselves while constrained maximizers can. Indeed, Edward McClennen formulates this difference as an empirical hypothesis: One might then reasonably expect to see this more efficient mode of dynamic decision-making drive out more costly precommitment and enforcement methods, and this through nothing more than what economists like to describe as the ordinary competitive process (Danielson (1998b) quoting an earlier version of McClennen (1998)).
However, to our knowledge, there have been no empirical tests of RGT; its defenders have favored theoretical arguments, augmented by science fiction thought experiments (Kavka, 1983; Viminitz, 1996) or computer simulations (Danielson, 1992, 1998c, 2001) to show that (variations on) constrained maximization is possible. 2.2. EGT: Too many solutions and norms With the RGT solution before us, we can introduce EGT as a critical alternative. First, a concession: according to orthodox game theory, and evolutionary
223
game theorists agree, in a one-shot PD, the noncooperative, non-optimal
outcome is the only equilibrium, so it is the solution of the game. That is, EGT concedes the point to RGT: EGT cannot provide a ‘‘cooperative solution’’ to the oneshot PD. Instead, it casts doubts on RGT’s proffered ‘‘solution’’ in two ways. First, EGT shifts attention from the compliance problem why should rational agents cooperate to the equilibrium selection problem. The first side of this is trivial; according to EGT, agents end up in an equilibrium, so there is no need further to secure their compliance to that outcome. As we just saw, the only equilibrium in the single-play PD is non-cooperative. So one needs to change the game, in this case, by repeating it, to get a more cooperative equilibrium. The new game provides opportunities for players to sanction non-cooperators; this new behavior is included in the equilibrium strategies. So again, there is no need to secure compliance to an out-of-equilibrium solution. The second part focuses on equilibrium selection. Once the game is repeated, there are many indeed infinitely many ways to cooperate. This is the so-called ‘‘folk theorem’’ of repeated games; every cooperative outcome can be reached by a set of equilibrium strategies.6 So while repeating the game solves the compliance problem, by moving cooperative interaction into equilibrium, it introduces a real problem of alternative cooperative norms. In particular, the prominence of particular cooperative strategies, such as tit-for-tat, is due mainly to the salience of research such as Axelrod (1984) rather than to any uniqueness of the strategies themselves (Binmore, 1998b). How does this result cast doubt on Gauthier’s RGT solution? Indeed, the multiplicity of solutions in the repeated game looks like an embarrassment for the EGT camp, compared to RGT’s unique solution. The answer is that the EGT approach sees the ‘‘unique’’ RGT outcome as an illusion, due to a failure of analysis. The mechanism for interaction in the repeated game is transparent: well-understood finite-state automata provide complete models of the players and generate the great complexity of possible equilibria (Binmore, 1992). Contrast the mechanism of constrained maximization. Players somehow keep themselves from choosing their preferred strategy when they somehow know that others are doing
224
PETER DANIELSON
something similar. Each clause hides complexity parallel to the alternative norms in the repeated game case. A full analysis would need to consider constrained maximizers of type X, who cooperate with constrained maximizers of type W and X perhaps, etc. It is difficult to unpack these strategies in the limited space game and cognitive of the one-shot game, yet when this expansion is attempted, the results are similar to the folk theorem (Danielson, 1998a, 2002). 2.3. Norms and ethics We now turn to the alternative account: evolutionary game theory. From the perspective of ethics, evolutionary game theory’s appeal to moral norms faces two problems. First, EGT needs to address the problem of moral freedom: how can shared social norms carry the normative burden that ethics requires? Second, can externally motivating norms explain our moral commitments? The answer to the first question is easy: norms do not carry the entire normative burden. Deploying a problem reduction strategy, EGT divides the burden between existing norms and recommendations for new norms. EGT is thus another two level theory, as discussed in note 1. A moral agent is both constrained and free; these contrasting descriptions focus on two levels of the EGT account. Moral norms constrain those subject to them, that is, moral agents. Our norms constrain our choice of actions directly and our reactions to other’s choices and our psychological reactions indirectly. But while the current equilibrium selects an option, it doesn’t eliminate the alternatives. Moral agents are free at this second level to prefer alternatives to the current norm, and to argue and politic for a change of norms. This freedom is not unlimited; the current norms still determine what is feasible, provide the baseline for bargaining for new norms, and, more controversially, the values with which alternatives are weighed (Binmore, 2004). The answer to the second question is more controversial. Binmore (2004), at one extreme, insists that the assumptions of rational choice theory suffice; Gintis (2000b) argues for a place for ‘‘strong reciprocity’’. The role of commitment remains an important open question for the evolutionary game theory approach, as does the weight given to theoretical vs. experimental evidence (Sugden, 2001).
3. Practical issues: Exploring real norms Having argued for EGT on theoretical grounds, we turn to the practice of applied ethics further to address the issues between the two approaches. At first look, this should be simple, as the difference between the two approaches appears to be large. According to RGT, ethics is more like a common logic or mathematics of social interaction, while according to EGT, local social norms play a greater role. This difference should be testable. However, teasing out how people make moral judgments is far from simple. We consider two cases that illustrate how rationalist and evolutionary game theories suggest quite different research methods for applied ethics and support the EGT approach with its focus on actual social norms. 3.1. Misleading intentions and mistaken norms The first case is a puzzle posed by Canada’s voluntary blood donation system. Plagued by low donation rates, the Canadian Blood Service recently commissioned a survey. The survey data suggests that while only 4% of Canadians actually give blood, about ¼ ‘‘intend’’ to donate and most Canadians believe that about ¼ actually do give blood (Canadian Blood Services, 2004). The RGT approach suggests an immediate interpretation: donors are in a prisoner’s dilemma where signaling false intentions to donate is a strategy to ‘‘sucker’’ others to cooperate (by donating). While this interpretation is plausible, it also tends to undercut a distinctive claim of RGT. If intentions can be so readily falsified, intentions to choose disadvantageous alternatives seem a poor foundation for ethics. We suggest that an EGT approach leads to a more fruitful approach. The fact that so many Canadians think so many Canadians will donate suggests an equilibrium based on poor information. Similar structures have been identified in the case of norms supporting heavy drinking on college campuses in the U.S. (Perkins and Berkowitz, 1986). So a normsbased approach leads us to ask: would this equilibrium persist were potential donors better informed about actual donation rates? A further question comes up on the norm-based EGT approach: how will behavior change given new information? Will donation rise or fall if people discover that more or less are donating? This should depend on the
PLAYING WITH ETHICS
distribution of norms and principles in the population. Crudely, altruists should tend to give more when informed on the low actual rate; reciprocators will find the low rate unfair and give less. Therefore, the norms-based EGT approach suggests that we need to find out more about the range and content of the norms and principles in the population. This is an empirical issue, not one settled by appeal to theory. (Indeed, in the case at hand, the policy that all blood donation is altruistic impedes the investigation of the variety of real mechanisms at work.)
3.2. An instrument to study norms Turning to a second case, our research group has recently designed and deployed a web-based survey tool aimed to provide in-depth information about norms and moral decision-making. There is a tension in our design between motivating participants by offering real moral issues about real and significant biotechnologies vs. the methodological need to construct more fictionalized experimental scenarios. We have tended to emphasize the former (Danielson et al., 2004). Our first survey is based on the history of the discovery and treatment of the genetic disorder ß-Thalassaemia in Cyprus (Bornik and Dowlatabadi, 2004) in the 20th century. Our web-survey instrument presents a set of 12 policy and personal decisions, allowing participants to consult with a series of advisors a medical scientist, a policy expert, an ethics expert, and pro and con advocates. (Readers inclined to take the survey are invited to do so now at http://yourviews.ubc.ca before they read the details to follow.) This project is nested as part of two larger research projects. For our Norms Evolving in Response to Dilemmas (NERD) research group, this is the first of a series of parallel surveys on salmon genomics and forestry genomics, which will allow us to address comparative issues across human health, food, and environment, which we will ignore here. For the Democracy, Ethics and Genomics project of which NERD is a part, other groups use methods focus groups and deliberative assemblies that control very little for social structure. Our online survey can control for social structure: we isolate our participants, allowing them access to others only via our panel of advisors and the feedback mechanism we will now describe.7 Obvi-
225
ously, we cannot control for participants’ access to social norms. Instead, we manipulated one parameter whether participants received feedback about how others answered the questions and consulted advisors. Throughout the survey, all participants address the same questions and have links to the same fictional advisor. Half of the participants also receive aggregated feedback information on how other participants (in their group) answered each question and which advisors they consulted; the other half does not receive feedback. We conjecture that such information will affect the way respondents in the feedback group answer the questions (Danielson et al., 2004: 6). Two of our fifteen hypotheses are directly relevant to the central issue of moral norms: N1: There will be a significant difference between participants’ answers with group feedback and without. N6: There will be a correlation between participants who receive feedback and their frequency of advice-seeking (Danielson et al., 2004: Appendix 1). The results of our first experimental survey were mixed. N1 was not confirmed but N6 was. That is, the answers were not significantly different between the feedback and non-feedback groups, but the nonfeedback group spent significantly more time consulting the advisors. (See Figure 2; the data summarizes approximately 30,000 data points ¼ 500 participants · 12 questions · 5 advisors). We conclude that feedback, a surrogate for the role of social norms, can have an effect even on responses to wellpublicized ethical issues. We look forward to our comparative tests with food and environmental issues further to refine these initial results. This research instrument is designed from an EGT perspective. It provides means to test the influence of social norms accessed in various ways: directly via feedback and indirectly via the advisors, especially the less reasoned Yes and No advocates. So, while not designed as a test of evolutionary vs. rationalist game theory, these experiments do inform the debate. So far, they point to a significant, though small, difference attributable to the norm analogue of social feedback.
226
PETER DANIELSON
Figure 2. Evidence for a Norm.
4. Conclusion We structured this paper by modeling moral seriousness as constraint and focusing on the play of alternatives for rational and moral agents. This served to frame our discussion of two ways in which game theory has been applied to ethics: rationalist and evolutionary game theories. We argued that the evolutionary game theory approach was likely to be more productive for our research program of modeling ethical mechanisms, appealing to both theoretical and practical considerations. Theoretically, EGT focuses our attention on well-understood social mechanisms and their cognitive complements, while RGT is committed to more extreme and poorly understood mechanisms. Practically, EGT informs applied ethics research by framing good research problems and suggesting new instruments to advance our understanding of our social norms.
decision’’. More generally, two-level accounts of ethics tend in this direction. Hare’s (1981) higher level is for philosophers; ‘‘people in ordinary life’’ remain at the lower level. In welcome contrast Rawls (1971) democratizes his account’s higher levels, by explicitly mapping them onto the U.S. constitutional structure. 2 See Varner (1995), who discusses where to broach radical critique of existing norms. 3 Witness Singer’s (2001) defense of bestiality and the attempts to exclude the controversial applied ethicist from Germany (Danielson and MacDonald, 1996). 4 We will focus exclusively on the RGT and EGT projections here, making no claim that our accounts apply to the broader philosophical traditions. 5 This is but one of several ways to contrast the approaches; Danielson (2004) takes a different approach; Skyrms (1996) is an excellent and the most accessible introduction to some of the differences. 6 Indeed, Binmore calls the folk theorem ‘‘the most significant insight available to political philosophy’’ (Binmore, 2004: 81). 7 Of course, we cannot control the social setting in which participants access the web, nor the sites they visit while taking our survey.
Notes * Thanks to Fabio Paglieri, Rana Ahmad, and Patrick Lewis for comments on a draft, to the NERD team for contributions to the research in section 3.2, and to the anonymous referees for their excellent suggestions. This research is funded by SSHRC and Genome Canada/BC. 1 Parfit (1984: 451452) writes that his conclusion ‘‘should if possible be concealed from those who will decide whether we increase our use of nuclear energy. These people know that the Risky Policy might cause catastrophes in the further future. It would be better if these people believe, falsely, that the choice of the Risky Policy would be against the interests of the people killed by such a catastrophe. If they have this false belief, they would be more likely to make the right
References Axelrod, R. M.: 1984, The Evolution of Cooperation, New York: Basic Books. Binmore, K.: 1994, Game Theory and the Social Contract: Playing Fair, Cambridge, MA: MIT Press. Binmore, K.: 1998a, Game Theory and the Social Contract: Just Playing, Cambridge, MA: MIT Press. Binmore, K.: 1998b, ‘Review of The Complexity of Cooperation by Robert Axelrod’, The Journal of Artificial Societies and Social Simulation 1. Binmore, K.: 2004, Natural Justice, Cambridge, MA: MIT Press.
PLAYING WITH ETHICS
Binmore, K.: 1992, Fun and Games: a Text on Game Theory, Lexington, MA: D.C. Heath. Bornik, Z., Dowlatabadi, H.: 2004, ‘The Interplay of Technological Change and Social Norms: The Case of B-Thalassaemia in Cyprus’, Working Paper, Centre for Applied Ethics, University of British Columbia. Canadian Blood Services.: 2004, June 14, World Blood Donor Day survey underscores ‘‘reality gap’’; Canadians exaggerate intentions to donate, overestimate the number of blood donors in Canada, Retrieved http://www.bloodservices.ca/centreapps/internet/UW_V502_MainEngine.nsf/761d50021d7897 9185256af-700577ec6/68c8218a06a544ab85256eb000692942 ?OpenDocument Danielson, P.: 1992, Artificial Morality, London-New York: Routledge. Danielson, P.: 1998a, ‘Evolutionary Models of Cooperative Mechanisms: Artificial Morality and Genetic Programming’, in P. Danielson (ed.), Modeling Rationality, Morality, and Evolution, New York: Oxford University Press, vol. 7, pp. 423441. Danielson, P: 1998b, ‘Introduction’, in P. Danielson (ed.), Modeling Rationality, Morality, and Evolution, New York: Oxford University Press, vol. 7, pp. 39. Danielson, P.: 2001, ‘Which Games Should Constrained Maximizers Play?’, in C. Morris, A. Ripstein (eds.), Practical Rationality and Preference: Essays for David Gauthier, New York: Cambridge University Press. Danielson, P.: 2002, ‘Competition among Cooperators: Altruism and Reciprocity’, Proceedings of the National Academy of Sciences 99, 72377242. Danielson, P.: 2004, ‘Rationality and Evolution’, in A. R. Mele, P. Rawling (eds.), The Oxford Handbook of Rationality, New York: Oxford University Press, pp. 417437. Danielson, P. (ed.): 1998c, Modeling Rationality, Morality, and Evolution, New York: Oxford University Press. Danielson, P., Ahmad, R., Bornik, Z., Dowlatabadi, H., Levy, E.: 2004, Deep, Cheap, and Improvable: Dynamic Democratic Norms & the Ethics of Biotechnology. Paper presented at the American Philosophical Association/NSF Conference: Ethics and the Life Sciences, Delaware. Danielson, P., MacDonald, C.: 1996, ‘Singer’s Agenda for Practical Ethics’, Dialogue 35, 599610. Gauthier, D. P.: 1986, Morals by Agreement, Oxford-New York: Clarendon Press; Oxford University Press. Gintis, H.: 2000a, Game Theory Evolving: a Problem-centered Introduction to Modeling Strategic Behavior, Princeton, NJ: Princeton University Press. Gintis, H.: 2000b, ‘Strong Reciprocity and Human Sociality’, J. Theoretical Biology 206, 169179. Hampshire, S.: 1978, ‘Morality and Pessimism’, in S. Hampshire (ed.), Public and Private Morality, Princeton: Princeton University Press, pp. 6282. Hare, R. M.: 1981, Moral Reasoning: its Levels, Method and Point, Oxford: Clarendon Press.
227
Kavka, G. S.: 1983, ‘The Toxin Puzzle’, Analysis 43, 3336. McClennen, E.: 1990, Rationality and Dynamic Choice, Cambridge: Cambridge University Press. McClennen, E.: 1998, ‘Rationality and Rules’, in P. Danielson (ed.), Modeling Rationality, Morality, and Evolution, New York: Oxford University Press, pp. 1340. Parfit, D.: 1984, Reasons and Persons, Oxford: Clarendon Press. Perkins, H., Berkowitz, A.: 1986, ‘Perceiving the Community Norms of Alcohol Use Among Students: Some Research Implications for Campus Alcohol Education Programming’, International Journal of Addictions 21, 961976. Rawls, J.: 1971, A Theory of Justice, Cambridge, MA: Harvard University Press. Schrage, M.: 2000, Serious Play: How the World’s Best Companies Simulate to Innovate, Boston, MA: Harvard Business School Press. Sidgwick, H.: 1893, Methods of Ethics, 5th ed., London-New York: Macmillan. Singer, P.: 2001, Heavy Petting: Review of Dearest Pet: On Bestiality by Midas Dekkers, Retrieved 23 Aug 2004, from http://www.nerve.com/Opinions/Singer/heavyPetting/ main.asp. Skyrms, B.: 1996, Evolution of the Social Contract, CambridgeNew York: Cambridge University Press. Skyrms, B.: 2003, The Stag Hunt and the Evolution of Social Structure, Cambridge-New York: Cambridge University Press. Strauss, L.: 1952, Persecution and the Art of Writing, Glencoe, IL: Free Press. Sugden, R.: 1986, The Economics of Rights, Co-operation, and Welfare, Oxford-New York: B. Blackwell. Sugden, R.: 2001, ‘Ken Binmore’s Evolutionary Social Theory’, Economic Journal 111, 213243. Varner, G. E: 1995, ‘Can Animal Rights Activists be Environmentalists’, in D. Marietta, L. Embree (eds.), Environmental Ethics and Environmental Activism, Rowman & Littlefield, pp. 169201. Viminitz, P.: 1996, ‘No Place to Hide: Campbell’s and Danielson’s Solutions to Gauthier’s Coherence Problem’, Dialogue: Canadian Philosophical Review 35, 235240. Young, H. P.: 1998, Individual Strategy and Social Structure: an Evolutionary Theory of Institutions, Princeton, NJ: Princeton University Press.
W. Maurice Young Centre for Applied Ethics University of British Columbia Vancouver Canada E-mail: [email protected]