Category Archives: moral network mapping

structured moral pluralism (a proposal)

(New York) Isaiah Berlin recalled that the Russian novelists he read as boy shared with “the major figures [of philosophy], especially in the field of ethical and political thought,” a common “Platonic ideal.” This ideal implied,

In the first place that, as in the sciences, all genuine questions must have one true answer and one only, all the rest being necessarily errors; in the second place that there must be a dependable path towards the discovery of these truths; in the third place that the true answers, when found, must necessarily be compatible with one another and form a single whole, for one truth cannot be incompatible with another – that we knew a priori. This kind of omniscience was the solution of the cosmic jigsaw puzzle. In the case of morals, we could then conceive what the perfect life must be, founded as it would be on a correct understanding of the rules that governed the universe (2013, p. 4) .

This passage is a simplification of intellectual history (Berlin himself cites Vico, Herder, and others as opponents of the view that he attributes to “the major figures”), but he accurately describes one tendency. For some important thinkers, moral truths–if they exist at all–must form a single whole, like a completed jigsaw puzzle or like a mechanism in which some components support or drive others. Not only should the elements be compatible, but articulable reasons or arguments should connect them together. If you believe A, you should be able to say why in terms of B. If you believe A and B, but the two seem to conflict, then you should be able to resolve the conflict by adjusting the two principles.

By the way, you can hold this model of moral thought even if you doubt, given our cognitive and moral limits, that we will ever see the whole puzzle correctly. The truth may still be a coherent structure even if what we know is always partial and confused.

Another view is very different from this one. It is the theory that human beings have instinctive, affective reactions to situations. After we form those reactions, we may rationalize them with arguments, but our arguments are always insufficient to determine our reactions, and we are good at gerrymandering our general principles to fit what we want to conclude about specific cases. Thus our arguments do not explain our judgments. However, empirical psychologists can detect patterns in our various reactions, which suggest the existence of unconscious latent factors that do explain what we feel about cases. Those factors may not be mutually compatible, which is why we are often ambivalent or inconsistent. They may also vary from person to person. But they exist, and what we say about moral issues is inconsequential compared to this structure of latent factors (see, e.g., Haigt and Graham et al.).

This view could be correct, although I suspect it is partly an artifact of the research methods. To the extent that it is true, it denies the value of moral deliberation, which is a fundamental obligation in the tradition that Berlin calls “Platonic.” Moral positions, Haidt writes, are “nearly impregnable to arguments from outsiders.” That implies an answer to the question that opens the Federalist Papers–“whether societies of men are really capable or not of establishing good government from reflection and choice, or whether they are forever destined to depend for their political constitutions on accident and force.” If latent factors determine responses, then we are destined to depend on accident. I hope that is not the case.

Berlin famously dissented from the “Platonic” view of morality and developed a version of pluralism. There are the main elements of his position:

  1. “There is a world of objective values” (p. 11). In other words, some things really are valuable. It is wrong to deny an actual value, such as freedom or equality, or to add something to the list of values that doesn’t merit inclusion. In short, there can be a right or a wrong answer to the question whether something (e.g., love, war, desire, loyalty) is a good. This is different from Moral Foundations theory, which presumes that we must value whatever we value.
  2. But the genuine “values can clash – that is why civilisations are incompatible. They can be incompatible between cultures, or groups in the same culture, or between you and me” (p. 12).
  3. Because of the nature of morality and/or human nature, there is no possible world inhabited by human beings in which all the goods are perfectly compatible. “These collisions of values are of the essence of what they are and what we are. … The notion of the perfect whole, the ultimate solution, in which all good things coexist, seems to me to be not merely unattainable – that is a truism – but conceptually incoherent; I do not know what is meant by a harmony of this kind. Some among the Great Goods cannot live together” (pp. 13-14).
  4. The misguided effort to harmonize all worthy values into one structure is a dangerous illusion (p. 15), or even “the road to inhumanity” (pp. 19-20), because it justifies the imposition of moral beliefs on others without compromises.

I am basically pluralist, but I would alter Berlin’s view in one important respect. He seems to assume a list of fully distinct and potentially incompatible goods. I observe that people make connections among some of their own ideas. They say that one value implies, or supports, or resembles another value in various respects.

These structures seem to me to have merit. Connecting two ideas means giving a reason for each of them, because now they hang together. We ought to reason in order to live an examined life and to deliberate with other people. We are prone to very grave limitations and biases if we merely rely on our instinctive reactions to moral situations, taken one at a time, or if we allow latent factors to determine our reactions. We should struggle to put our ideas together into explicit structures and should present portions of those structures to other human beings for inspection and critique. That is just an idiosyncratic way of saying that we must reason together about values. Reasoning does not mean endorsing various Great Goods, one at a time, but rather connecting each idea to another idea.

This view is still compatible with Berlin’s pluralism, for two important reasons. First, the structure of moral ideas that each of us gradually builds and amends may contain incompatible values. Each of us can be a pluralist, even as we attempt to connect many of our own ideas into networks. Our networks can contain gaps and loose links and can reflect tradeoffs. Second, is it likely that even human beings who strive to develop the best possible structures of moral ideas will never produce the same structures. That is because moral reflection is deeply dependent on local experience and on conversations with concrete other people, each of whom is affected by her own conditions. So we will forever disagree. In contrast to the image of a “cosmic jigsaw puzzle” that we are all working together to complete, I’d propose a great web of loosely connected ideas that we are all perpetually creating and linking together.

See also: 10 theses about ethics, in network termsJonathan Haidt’s six foundations of moralityan alternative to Moral Foundations Theory; and everyone unique, all connected.


Berlin, Isaiah. The crooked timber of humanity: Chapters in the history of ideas. Princeton University Press, 2013.

Jonathan Haidt, The Righteous Mind: Why Good People are Divided by Politics and Religion (New York: Pantheon, 2012)

Jesse Graham, Brian Nosek, Jonathan Haidt, Ravi Iyer, Spassena Koleva, and Peter H.Ditto, “Mapping the Moral Domain,” Journal of Personality and Social Psychology, Vol 101(2), Aug 2011, 366-385.

an alternative to Moral Foundations Theory

Jonathan Haidt’s Moral Foundations Theory is one of the most influential current approaches to moral psychology and it exemplifies certain assumptions that are pervasive in psychology more generally. I have been working lately with 18 friends and colleagues to “map” their moral views in a very different way, driven by different assumptions. As part of this small pilot project, I gave the 18 participants Haidt et al’s, Moral Foundations Questionnaire. Although my sample is small and non-representative, I am interested in the contrasting results that the two methods yield.

Haidt’s underlying assumptions are that people form judgments about moral issues, but these are often gut reactions. The reasons that people give for their judgments are post-hoc rationalizations (Haidt 2012, pp. 27-51; Swidler 2001, pp. 147-8; Thiele 2006). “Individuals are often unable to access the causes of their moral judgments” (Graham, Nosek, Haidt, Iyer, Koleva, & Ditto 2011, p. 368). Hence moral psychologists are most interested in unobserved mental phenomena that can explain our observable statements and actions.

Haidt et al ask their research subjects multiple-choice questions about moral topics. Once they have collected responses from many subjects, they use factor analysis to find latent variables that can explain the variance in the answers. (Latent variables have been “so useful … that they pervade … psychology and the social sciences” [Bollen, 2002, p. 606]). The variables that are thereby revealed are treated as real psychological phenomena, even though the research subjects may not be aware of them. Haidt and colleagues consider whether each factor names a psychological instinct or emotion that 1) would have value for evolving homo sapiens, so that our ancestors would have developed an inborn tendency to embrace it, and 2) are found in many cultures around the world. Now bearing names like “care” and “fairness,” these factors become candidates for moral “foundations.”

Because Haidt’s method generates a small number of factors, he concludes that people can be classified into large moral groups (such as American liberals and conservatives) whose shared premises determine their opinions about concrete matters like abortion and smoking. “Each matrix provides a complete, unified, and emotionally compelling worldview” (Haidt 2012, p. 107). In this respect, Haidt’s Moral Foundations theory bears a striking similarity to Rawls’ notion of a “comprehensive doctrine” that “organizes and characterizes recognized values to that they are compatible with each other and express an intelligible view of the world.”

In contrast, I have followed these steps:

  1. I recruited people I knew. These relationships, although various, probably influenced the results. I don’t entirely see that as a limitation.
  2. I asked each participant to answer three open-ended questions: “Please briefly state principles that you aspire to live by.” “Please briefly state truths about life or the world that you believe and that relate to your important choices in life.” “Please briefly state methods that you believe are important and valid for making moral or ethical decisions.”
  3. I interviewed them, one at time. I began by showing each respondent her own responses to the the survey, distributed randomly as dots on a plane. I asked them to link ideas that seemed closely related. When they made links, I asked them to explain the connections, which often (not always) took the form of reasons: “I believe this because of that.” As we talked, I encouraged them to add ideas that had come up during their explanations. I also gently asked whether some of their ideas implied others yet unstated; but I encouraged them to resist my suggestions, and often they did. The result was a network map for each participant with a mean of 20.7 ideas, almost all of which they had chosen to connect together, rather than leaving ideas isolated.
  4. We jointly moved the nodes of these networks around so that they clustered in meaningful ways. Often the clusters would be about topics like intimate relationships, views of social justice, or limitations and constraints.
  5. I put all their network maps on one plane and encouraged them to link to each others’ ideas if they saw connections. That process continues right now, but the total number of links proposed by my 18 participants has now reached 1,283.
  6. I have loosely classified their ideas under 30 headings (Autonomy, Authenticity/ integrity/purpose, Balance/tradeoffs, Everyone’s different but everyone contributes, Community, Context, Creativity/making meaning, Deliberative values, Difficulty of being good, Don’t hurt others, Emotion, Family, Fairness/equity, Flexibility, God, Intrinsic value of life, Justice, Life is limited, Maturity/experience, Modesty, No God, Optimism, Peace/stability, Rationality/critical thinking, Serve/help others, Relationships, Skepticism/human cognitive limitations, Striving, Tradition, Virtues). Note that some of these categories resemble Moral Foundations, but several do not. The ones that don’t tend to be more “meta”–about how to form moral opinions.

My assumptions are that people can say interesting and meaningful things in response to open-ended questions about moral philosophy; that much is lost if you try to categorize these ideas too quickly, because the subtleties matter; and that a person not only has separate beliefs but also explicit reasons that connect these beliefs into larger structures.

Since I also gave participants the Moral Foundations Questionnaire, I am able to say some things about the group from that perspective. This graph shows the group means and the range for their scores on the five Moral Foundations scales. For comparison, the average responses of politically moderate Americans are 20.2, 20.5, 16.0, 16.5, and 12.6. That means that my group is more concerned about harm/care and fairness/reciprocity than most Americans, and not far from average on other Foundations. But there is also a lot of diversity within the group. Two of my respondents scored 5 out of 35 on the purity scale, and two scored 20 or higher. The range was likewise from 6 to 28 on the in-group/loyalty scale.


You might think that this diversity would somehow be reflected in the respondents’ maps of their own explicit moral ideas and connections. But I see no particular relationships. For instance, one of the people who rated purity considerations as important–a self-described liberal Catholic–produced a map that clustered around virtues of moral curiosity and openness, friendship and love, and a central cluster about justice in institutions. She volunteered no thoughts about purity at all.

This respondent scored 20 on the purity scale. A different person (self described as an atheist liberal) scored 9 on that scale. But they chose to connect their respective networks through shared ideas about humility, deliberation, and justice.

The whole group did not divide into clusters with distinct worldviews but overlapped a great deal. To preserve privacy, I show an intentionally tiny picture of the current group’s map that reveals its general shape. There are no signs of separate blocs, even though respondents did vary a lot on some of the “Foundations” scales.


A single-word node that appears in five different people’s networks is “humility.” It also ranks fourth out 375 ideas in closeness and betweenness centrality (two different measures of importance in a network). It is an example of a unifying idea for this group.

Many of the ideas that people proposed have to do with deliberative values: interacting with other people, learning from them, forming relationships, and trying to improve yourself in relation to others. Those are not really options on the Moral Foundations Questionnaire. They are important virtues if we hold explicit moral ideas and reasons and can improve them. They are not important virtues, however, if we are driven by unrecognized latent factors.

One way to compare the two methods would be to ask which one is better able to predict human behavior. That is an empirical question, but a complex one because many different kinds of behavior might be treated as outcomes. In any case, it’s not the only way to compare the two methods. They also have different purposes. Moral Foundations is descriptive and perhaps diagnostic–helping us to understand why we disagree. The method that I am developing is more therapeutic, in the original sense: designed to help us to reflect on our own ideas with other people we know, so that we can improve.

[References: Bollen, Kenneth A. 2002. Latent Variables in Psychology and the Social Sciences. Annual Review of Psychology, vol. 53, 605-634; Graham, Jesse, Nosek, Brian A., Haidt, Jonathan, Iyer, Ravi, Koleva. Spassena, & Ditto, Peter H. 2011. Mapping the Moral Domain. Journal of Personality and Social Psychology, 101:2; Haidt, Jonathan. 2012. The Righteous Mind: Why Good People Are Divided by Politics and Religion. New York: Vintage; Swidler, Ann. 2001. Talk of Love: How Culture Matters. Chicago: University of Chicago Press; Thiele, Leslie Paul. 2006. The Heart of Judgment: Practical Wisdom, Neuroscience, and Narrative Cambridge University Press.]

10 theses about ethics, in network terms

  1. People hold many morally relevant opinions, some concrete and particular, some abstract and general, some tentative and others categorical.
  2. People see connections–usually logical or empirical relationships–between some pairs of their own opinions and can link all of their opinions into one network. (Note: these first two theses are empirical, in that I have now “mapped” several dozen students’ or colleagues’ moral worldviews, and each person has connected all of his or her numerous moral ideas into a single, connected network. However, this is a smallish number of people who hardly reflect the world’s diversity.)
  3. Explicit moral argumentation takes the form of citing relevant moral ideas and explaining the links among them.
  4. The network structure of a person’s moral ideas is important. For instance, some ideas may be particularly central to the network or distant from each other. These properties affect our conclusions and behaviors. (Note: this is an empirical thesis for which I do not yet have adequate data. There are at least two rival theses. If people reason like classical utilitarians or rather simplistic Kantians, then they consistently apply one algorithm in all cases, and network analysis is irrelevant. Network analysis is also irrelevant if people make moral judgments because of unconscious assumptions and then rationalize them post hoc by inventing reasons.)
  5. Not all of our ideas are clearly defined, and many of the connections that we see among our ideas are not logically or empirically rigorous arguments. They are loose empirical generalizations or rough implications.
  6. It is better to have a large, complex map than a simple one that would meet stricter tests of logical and empirical rigor and clarity. It is better to preserve most of a typical person’s network because each idea and connection captures valid experiences and serves as a hedge against self-interest and fanaticism. The emergent social world is so complex that human beings, with our cognitive limits, cannot develop adequate networks of moral ideas that are clear and rigorous.
  7. Our ideas are not individual; they are relational. We hold ideas and make connections because of what others have proposed, asked, made salient, or provoked from us. A person’s moral map at a given moment is a piece of a community’s constantly evolving map.
  8. We begin with the moral ideas and connections that we are taught by our community and culture. We cannot be blamed (or praised) for their content. But we are responsible for interacting responsively with people who have had different experiences. Therefore, discursive virtues are paramount.
  9. Discursive virtues can be defined in network terms. For instance, a person whose network is centralized around one nonnegotiable idea cannot deliberate, and neither can a person whose ideas are disconnected.  If two people interact but their networks remain unchanged, that is a sign of unresponsiveness.
  10. It is a worthwhile exercise to map one’s own current moral ideas as a network, reflect on both its content and its form, and interact with others who do the same.

the advantages and drawbacks of precision in ethics

subject3I like to ask people to state their own beliefs that are relevant to ethics and then draw connections among those ideas to create networks that represent their moral worldviews. I put people (students and others) in dialogue with each other, invite them to explain their networks to peers, and watch connections form.

Usually the ideas that people propose are not precise. In explaining what we believe, we don’t employ many terms that we could define with necessary and sufficient conditions, nor do we often use quantifiers like “all” or “exactly one.” The connections we detect among our ideas are rarely logical inferences. They are looser links: resemblances, rough implications, empirical generalizations.

One impulse is to strive for as much precision as possible. That is a fundamental goal of analytic moral philosophy and it has significant merit. If someone proposed, “We should strive to improve everyone’s lives,” I would join mainstream analytic philosophers in requesting more clarity. Does that mean maximizing net human welfare? Does “welfare” mean happiness, satisfaction, or objective well-being? Does it trade off against freedom and autonomy? Does “everyone” mean all currently living human beings? (What about future generations?) Does “strive” mean actually maximize net welfare, or have a generally beneficent attitude toward others? These are valid and hard questions.

On the other hand, if the goal is descriptive moral psychology, it is a mistake to ask for that level of precision. We all hold–and are motivated by–rougher moral ideas and looser connections than could pass muster with an analytical philosopher. If you want to know what people believe, you must model those ideas and relationships as well as the clear ones. If you encourage people to map out many of their ideas and relationships, they will produce complex and elaborate networks that are useful for representing their mentalities and for provoking reflection.

That still leaves the normative question: how much precision should each of us strive for? I would say some but not too much. One of my favorite quotes is from Bernard Williams, in Ethics and the Limits of Philosophy (1985, p. 117):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.”

I’d expand that remark as follows: Through direct and vicarious experience, we build up collections of moral ideas that give our lives meaning and restrain our basest instincts. We also connect our ideas; we say that we believe A because it seems somehow related to B. If we must pass all these ideas and connections through a screen for clarity, precision, and inferential rigor, most will have to go. That will leave us with less meaning and less constraint against mere inclination and will.

Seeking clarity can illuminate. It can, for instance, force us to disaggregate a vague idea into a set of related ideas that are worth seeing on their own. Or it can reveal gaps and tradeoffs that deserve consideration. Formal philosophy is also useful for developing specific ideas that are clear and precise and that relate to one another logically.

However, it is a false dream that we can convert our entire networks of moral ideas into structures of clearly defined concepts and implications. Even the best moral arguments carry just a short distance–from a premise to a conclusion, or maybe as far as another conclusion or two, but not all the way across the domain of the moral. It is good to have a dense, complex, and expansive network of ideas that draws on experience and demands constant reflection and reevaluation, even if its components are a bit vague and the links are hard to articulate. Better that than a crystalline chain of reasons that connects just a few ideas and leaves us otherwise free to be selfish or fanatical.

Selim Berker on moral coherence

In “Coherentism via Graphs,”[i] Selim Berker begins to work out a theory of the coherence of a person’s beliefs in terms of its network properties. Consider these two diagrams (A and B) borrowed from his article, both of which depict the beliefs that an individual holds at a given time. If one beliefs supports another, they are linked with an arrow.

Screen Shot 2016-02-07 at 12.47.04 PM

Both diagrams show an individual holding three connected and mutually consistent beliefs. Thus traditional methods of measuring coherence can’t differentiate between these two structures. However, Graph A is pretty obviously problematic. It involves an infinite regress—or what has been called, since ancient times, “circular reasoning.” Graph B is far more persuasive. If someone holds beliefs that are connected as in B, the result looks like a meaningfully coherent view. If you find coherence relevant to justification, then you will have a reason to think that the beliefs in B are justified—a reason that is absent in A.

Berker also proposes a subtler but more decisive reason that B is better than A. Below I show A again, now with the component beliefs labeled as P, Q, and R. If the law of contraposition holds, than A implies another graph, A’, that is its exact opposite. A’  includes beliefs -P, -Q, and -R, and the arrows point in the reverse direction.

Screen Shot 2016-02-07 at 12.48.37 PM

But that means that if belief P is justified because it is part of a coherent system of beliefs, then the same must be true of -P, which is absurd.[ii]

The overall point is that coherence is a property of the network structure of beliefs. That should be interesting to coherentists, who argue that what justifies any given belief just is its place in a coherent system. But it should also be interesting to foundationalists, who believe that some beliefs are justified independently of their relations to other ideas. Foundationalists still recognize that many, if not most, of our beliefs are justified by how they are connected to other beliefs. Thus, even though they believe in foundations, they still need an account of what makes a worldview coherent.

I have been developing a similar view, with a narrower application to moral thought (and without Berker’s deep grasp of current epistemology). I am motivated, first, by the sense that what makes a moral worldview impressively coherent cannot be seen without diagramming its whole structure. Imagine, for instance, a person who holds two major moral beliefs: “Never lie” and “Do not eat meat.” Assume that this person has not found or seen any particular connection between these two main ideas.

Screen Shot 2016-02-07 at 12.50.10 PM

His or her set of maxims is perfectly consistent: there is no contradiction between any two nodes. And every idea has a connection to another. But if we wanted to judge the coherence of this worldview, we would not be satisfied with knowing the proportion of the components that were consistent and directly connected. It would matter that the person holds two separate clusters of ideas—two hubs with spokes. This person’s network is fairly coherent insofar as it is organized into clusters rather than being completely scattered; but it would be more coherent if the two clusters interconnected via large integrating ideas. You can’t see the problem without diagramming the structure.

I also have another motivation for wanting to explore moral worldviews and political ideologies as networks of beliefs. In moral philosophy and political theory, constructed systems are very prominent. Although diverse in many respects, such systems share the feature that they could be diagrammed neatly and parsimoniously. In utilitarianism, the principle of utility is the hub, and every valid moral judgment is a spoke. That theory is so simple that to diagram it would be trivial. Kantianism centers on several connected principles, and Aristotelian, Thomist, and Marxist views are perhaps more complicated still. But in every case, a network diagram of the theory would be organized and regular enough that the whole could be conveyed concisely in words.

In contrast, my own moral worldview has accumulated over nearly half century as I have taken aboard various moral ideas that I’ve found intuitive (or even compelling) and have noticed connections among them. My network is now very large and not terribly well organized. A narrative description of it would have to be lengthy and rambling. Many of my moral beliefs are nowhere near each other in a network that sprawls widely and clusters around many centers.

I suspect this condition is fairly typical. No doubt, individuals differ in how large, how complex, and how organized their moral worldviews have become, but a truly organized structure is rare. (I have asked a total of about 60 students and colleagues to diagram their own views, and only one of the 60 gave me a network that could be concisely summarized.) That means that such constructed systems as Kantianism and utilitarianism are remote from most people’s moral psychology.

Further, I think that having a loosely organized but large and connected network is a sign of moral maturity. It is a Good Thing. That is obviously a substantive moral judgment, not a self-evident proposition. It arises from a certain view of liberalism that would take me more than a blog post to elucidate. But the essential principle is that we ought to be responsive to other people’s moral experiences.

Berker includes experiences as well as beliefs in his network-diagrams of people’s worldviews.[iii] In science, it should not matter who has the experience. An experience of a natural phenomenon is supposed to be replicable; you, too, can climb the Leaning Tower and repeat Galileo’s experiment. But in the moral domain, experience is not replicable or subject-neutral in the same way. Since I am a man, I cannot experience having been a woman my whole life so far. Thus vicarious experiences are essential to moral development.

If we are responsive, we will accumulate sprawling and random-looking networks of moral beliefs as we interact with diverse other people. These networks can be usefully analyzed with the techniques developed for analyzing large biological and social networks. It will be illuminating to look for clusters and gaps and for nodes that are more central than average in the structure as a whole. The coherence of such a network is not a matter of the proportion of the beliefs that are consistent with each other. Its coherence can better be evaluated with the kinds of metrics we use to assess the size, connectedness, density, centralization, and clustering of the complex networks that accumulate in nature.

On the other hand, if someone adopts a moral view that could be diagrammed as a simple, organized structure, he has not been responsive to others so far and he will be hard pressed to incorporate their experiences in the future. At the extreme, his simple graph is a sign of fanaticism.

See also: envisioning morality as a network; it’s not just what you think, but how your thoughts are organized; Stanley Cavell: morality as one way of living well; and ethical reasoning as a scale-free network (my first thoughts along these lines, from 2009).


[i] Berker, S. (2015), Coherentism via Graphs. Philosophical Issues, 25: 322–352. doi: 10.1111/phis.12052

[ii] “Coherence, we have been assuming, is a matter of the structure of support among a subject’s beliefs, experiences, and other justificatorily-relevant mental states at a given time.” But we can use directed hypergraphs (in mathematics, networks in which any of the nodes can be connected to any number of the other nodes by means of arrows) to represent all of those support relations. That is, we use directed hypergraphs to represent all of the relations that have a bearing on coherence. It follows that coherence is itself expressible as a graph-theoretic property of our directed hypergraphs (p. 339).

[iii] “Many theorists hold that a subject’s perceptual experiences are justificatorily relevant (in these sense that they either partially or entirely make it the case that the subject is justified in believing something).”