Let us posit that a person’s moral worldview is a network. The nodes are commitments, beliefs, principles, and other moral ideas. The connections are logical. But, since we are thinking about morality and not mathematics, our criteria for valid connections must be loosened. Two moral ideas are connected not only in cases when A implies B, but also if A sets a precedent for B, A is analogous to B, A is a compelling example of B, or A and B came together harmoniously in the life of a person we admire. We can examine any moral network map and ask whether it has desirable features as a whole. Improving our own map becomes a method of introspective self-improvement.

Many works in the philosophical tradition push us to ask whether all the nodes in our own maps are mutually consistent and whether all the connections are logically tight. But those are only two virtues of a moral worldview, and they are easily overrated. A moral monster could have a consistent network in which all his concrete beliefs followed from a few premises by adamantine chains of logic. This evil person might hold horrible premises, or his assumptions might be wonderful (e.g., “Freedom for everyone!”), yet his whole system could be fanatically simplistic. Although moral networks vary in quality, and we should strive to improve them, coherence is not their main virtue.

Simon Blackburn asks:

Is it likely that becoming more coherent is the only kind of improvement in my own sensibility that I can contemplate? No increases of imagination, sensitivity, or empathy? No cultivation of hidden joys or unlocking of virgin springs of contentment, no larger views or wider sympathies? This sounds like nonsense. And similarly the like view that the method must be entirely one of improving reflective equilibrium. Such suggestions appear to take the current stock of attitudes as a datum, with the problem only that of making them rub along properly with each other; whereas often real improvement must require an expansion and change of the stock.”¹

I can imagine someone responding: But wouldn’t an ideal moral worldview be perfectly coherent? Wouldn’t God’s moral ideas, for example, all fit together perfectly? I am not sure: it is possible that the whole moral truth happens not to be coherent, because we have developed into a “crooked timber” through contingent evolutionary and historical processes. But even if the ideal moral worldview is coherent, it does not follow that we make our actual beliefs better (closer to God’s, so to speak) by making them more consistent. The typical ways to promote consistency are to reject ideas that do not fit with others or to replace conflicting ideas with more general and abstract ones. But Bernard Williams objects:

Reflective criticism should basically go in a direction opposite to that encouraged by ethical theory. Theory looks characteristically for considerations that are very general and have as little distinctive content as possible, because it is trying to systematize and because it wants to represent as many reasons as possible as applications of other reasons. But critical reflection should look for as much shared understanding as it can find on any issue, and use any ethical material that, in the context of the reflective discussion, makes some sense and commands some loyalty. . … Theory typically uses the assumption that we probably have too many ethical ideas, some of which may well turn out to be mere prejudices. Our major problem now is actually that we have not too many but too few, and we need to cherish as many as we can.²

Both Blackburn and Williams hint that a moral network map would be better if it contained two challenging and worthy ideas that were in fruitful tension with each other than if it were made more consistent by deleting or papering over the tension.

In fact, we might connect two moral beliefs if, instead of A implying B, A contends with B in ways that we should take seriously. For example: “All human beings count for the same” and “I care most about my own loved ones” may both belong on my moral map, and they should be connected with an arrow that stands for “irresolvable tension.” That arrow is an important link in my network. To delete either belief would make the map worse, as would trying to make them both consistent with some higher principle, such as utilitarianism or Kantianism.

There is precedent for seeing a person’s “repertoire of beliefs” as a network and asking whether the whole network reflects virtues–without presuming that the only relevant virtues are coherence or that strict logical entailment is the only valid connection among beliefs. I take the phrase “repertoire of beliefs” from W.V. Quine and J.S. Ullian’s introductory textbook, The Web of Belief, which has a helpful title.³

The authors note that not every truth is “derivable … by self-evident steps from self- evident truths together with observations.” On the contrary, induction and hypothesis are essential components of any worthy web of belief. Quine and Ullian then propose five generic “virtues” of hyphotheses: conservatism (minimizing conflicts with previous beliefs), modesty, simplicity, generality, and refutability. These virtues can trade off–for example, more generality may be purchased at the price of less modesty–and therefore, assessing a whole Web of Belief is a matter of judgment.

The focus of Quine and Ullian is not moral thinking; it is science. They do address ethics in the very last pages of their book, but they do so, it seems to me, in an unsatisfactory manner. First, they hope that many moral disagreements can be replaced with scientific ones. “Given a liking for the sweet-sour or spicy, or for Pissarro or Beethoven or the welfare of orphans, we have then only to inculcate a belief that the thing in hand or the act in prospect will promote one of those desired ends.” In this sentence, the welfare of orphans (which I would call an obligation) plays the same role as a preference for spicy food; both are goals that individuals just happen to hold, and we can rationally investigate how to achieve them.

But surely morality requires more than reasoning about means, because, as Quine and Ullian note, reasonable people disagree “over abortion, euthanasia, compulsory birth control, eugenic sterilization, [and] capital punishment.” And, at a higher level of abstraction, we  disagree about things like how to define altruism and how much altruism is obligatory. The authors make an analogy to “esthetic matters”:

There is no disputing about tastes: no disputing in the sense in which one can dispute beliefs. There is only a training of tastes. . … What of the welfare of orphans as something to be desired for itself? Here again there is scope for training, but in this case the sort of training that we first think of is different in principle from training in aesthetic appreciation; it is reward and punishment, like the training of dogs.

That last sentence trivializes moral discourse, but Quine and Ullian add:

There is room also, however, for moral training of a kind similar to training in aesthetic appreciation. The technique there was the selective emphasis of some telling traits of the aesthetic object. Now altruism, similarly, can be encouraged by vividly depicting the suffering that might be relieved by the altruistic act, or the joy that might be conferred. For this kind of training to succeed there must already be some springs of sympathy to draw on, but there generally are such: and not just because of earlier moral training by reward and punishment, but–as Hume saw–because of sheer inherited instinct.

This won’t suffice either; it reduces all moral philosophy and other rigorous inquiries into ethics and justice (e.g., through history, fiction, and literary criticism) to aesthetic training. Although Quine and Ullian acknowledge the existence of complex, principled moral arguments, they really don’t explain how those arguments can be evaluated.

Yet they point in a useful general direction. A theory of the “Moral Web of Belief” would resemble their account of good scientific reasoning. It would posit virtues not of individual beliefs but of the whole web. It would distinguish between valid and invalid connections among beliefs, but it would not presume that the only worthy connections are logical entailments.

Another model is John L. Pollack’s Cognitive Carpentry.⁴ Like Quine and Ullian, Pollack presents an unsatisfactory account of ethics: “As I shall use the term, to live the good life is to live so that, insofar as possible, the situation one is in is likable. Practical cognition seeks to direct the agent in its pursuit of the good life.” Both Mussolini and Martin Luther King, Jr. used practical cognition; we must be able to say which one sought better states of affairs. But Pollack is useful, first, because he sees the value of “introspection”: assessing one’s overall web or map of beliefs. Second, he uses the methods of network analysis, putting beliefs on a literal graph and connecting them with arrows that indicate logical relationships. He sees that one can reasonably reach a conclusion because of the overall graph, not just a particular deduction. Third, he rightly relaxes the standard for what counts as an “adequate” relationship between two beliefs. He is looking for defeasible reasons, ones that are compelling without being deductively valid.

What would be the characteristics of a good moral web or moral network map? I don’t find adequate answers within moral theory itself, although I may have missed them. Another place to look is in graph theory and social network analysis. These fields provide insights about what makes strong networks among people and among components of physical systems. When we borrow the same criteria for the moral domain, I find the results attractive. In assessing a moral worldview, we would look for virtues like the following:

• Cohesion: Few ideas or clusters of ideas are orphaned. Instead, one can get from most of the beliefs to most of the others by at least one path. So, for example, I should not just happen to favor saving birds unless I can connect that preference to other ideas about conservation, nature, beauty–or something else.
• Density: There are many connections among the beliefs; lots of paths between any node in the map and many of the rest. So, for example, if I favor gay marriage and also heterosexual marriage, that ought be because I hold lots of interconnected ideas about commitment and love, equality, freedom of choice, respect, the role of the state, particular couples I know, and so on. If everything rests on one simple idea–such as equality–it is much weaker.
• Closure: If A is related to B, and B is related to C, then it’s better for A and C to directly relate. But A and B might go together comfortably, while A and C might be in creative tension. That is still closure, as long as we recognize the tension and its implications.
• Proper centrality: The nodes that are most connected to others (those that “carry the most traffic,” in network terms) are weighty and publicly defensible. They are not, for example, trivialities or idiosyncratic preferences. I would actually go further and argue that the central nodes ought always to be concerned with the three essential topics of human concern: truth; community or justice; and happiness or equanimity.
• Breadth: Taken together, the nodes cover a wide range of issues and concerns. If I have a dense and impressive network that says nothing about global justice, or nature, or the inner life, it is missing important domains.
• Variety: Because humanity is a “crooked timber,” and we human beings have built complex structures over thousands of years, no small set of assumptions will suffice. Tensions and even inconsistencies are better than fanaticism.

I am implying an ethical methodology: analyzing moral beliefs as a network. This method could be used for actual introspection–to analyze one’s own views critically and improve them. Then it would be a contribution to philosophy as a way of life, per Pierre Hadot or John M. Cooper. It could meanwhile be refined at a theoretical level by professional moral philosophers, who might apply insights from logic and epistemology: for instance, from the study of defeasable reasoning. The method could also be applied in other domains; for instance, literary critics could map the moral networks implied by literary works.

This method poses a challenge, I think, to some common methods of ethics. As noted already, it rejects consistency as the main virtue of systems of moral thought. It is thus an alternative to “reflective equilibrium,” the very widespread method that begins with moral intuitions and tries to make them more consistent. Or, if the method I propose counts as a form of “wide reflective equilibrium,” it is very wide indeed. It accommodates diverse types of beliefs (abstract and concrete, personal and impersonal, vague and precise) and it views the network as strong if some of the elements are in creative tension.

This approach would use thought-experiments sparingly and in a particular way. Google Scholar finds 1,530 articles that involve “trolley problems,” thought-experiments about an out-of-control trolley heading toward a junction and threatening to crush people. Should you actively intervene to make the trolley hit one person instead of the five it was going to kill? (But are you not then responsible for the one person’s death?) If you can only stop the trolley by throwing a heavy person in front of it, should you? The usual idea is that we will learn something fundamental about our general moral principles by considering such  extreme cases. For instance, we can either believe “Thou shalt not kill” or “Maximize the welfare of people,” but not both, because they imply different answers to trolley problems.

Concrete and vivid thought experiments enlist powerful psychological processes and are therefore persuasive about the cases they present. Sometimes, they are persuasive for good reasons, snapping us out of our usual self-serving assumptions. Consider, for instance, Nathan’s story of the rich man and the poor man in the Second Book of Samuel (as analyzed by Tamar Szabó Gendler).⁵ Nathan’s hypothetical case makes King David realize he has sinned when he recognizes the analogy between his own behavior and that of the rich man in the story. The moral effect on David is beneficial. But the mere fact that a story is persuasive tells us little, because (as Gendler shows) we can construct scenarios that persuade in all kinds of directions. So everything depends on how the story fits into the whole moral network.

I might include my own responses to trolley problems in my moral network map, but I would not be surprised to find my intuitions about different thought-experiments connected to general principles that conflict with each other. Further, I would expect these cases to be marginal on my map. I am not aiming for a systematic moral theory that might resemble, for example, utilitarianism or Kantian deontology. I am not expecting everything to derive from “considerations that are very general and have as little distinctive content as possible” (again quoting Williams). I am rather striving for a rich moral worldview. Stylized cases involving extreme outcomes will just find their places amid many real-world examples that matter more to me.

See also: using the full space of moral reasons, moral network mapping and literary criticism: a methodological proposal, all that matters is equanimity, community, and truth, mapping a moral network: Auden in 1939, and morality as a network (revisited).

¹Blackburn, “Securing the Nots,” in Walter Sinnott-Armstrong and Mark Timmons, eds., Moral Knowledge (New York, Oxford University Press, 1996), p. 95.
²Williams, Ethics and the Limits of Philosophy, (Cambridge, Mass: Harvard, 1985), p. 117.
³W.V. Quine and J.S. Ullian, The Web of Belief (McGraw-Hill Humanities/Social Sciences/Languages; 2nd edition, 1978)
⁴John L. Pollock, Cognitive Carpentry: A Blueprint for how to Build a Person, MIT Press, 1995.
⁵Gendler, “Philosophical Thought Experiments, Intuitions, and Cognitive Equilibrium,” Midwest Studies in Philosophy, 31 (2007), pp. 68-89