The Structure of a Worldview

Grab-bag of beliefs and values, or leaves on a decision tree?

Nov 15, 2024

All this thinking about worldviews and cultural drift has been deeply engaging. Between Robin Hanson questioning the rationality of trusting in our deepest cultural values, and reading Thomas Sowell, who very clearly articulates the anti-idealist position and bias in favor of tradition—I’ve been evaluating political and moral disagreements in a new light. This essay continues my exploration of worldviews and the “worldview space” by examining their potential structure, particularly how they form, influence other beliefs, and help to explain seemingly unrelated differences among groups of individuals. And I’ll once again tie this discussion back to the analogy with personality, and some relevant details from the Big 5 research program.

So, what axis of variation explains the maximum amount of relevant and salient differences in the worldviews of individuals within a given population and time period of interest? Or alternatively, since it’s not obvious that there’s always one dominant foundational divide, how could we identify a manageable set of largely independent factors which together explain a significant amount of salient differences across individuals in the context of interest.

Well, it really depends on what the context of interest is and what sorts of differences (variance) you want the concept of a worldview to explain. Maybe you want to deeply understand the key axes along which feminists differ1, or maybe you want to understand current divisions on issues in national politics, or maybe the divisions most relevant in the Western world over the entire post Enlightenment period, or maybe the differences which divide various cultures across all of recorded history, or maybe differences in the outcomes at the societal level (in terms of scientific progress, wealth production etc.) that a culture’s general position along certain axes predict. Or maybe something else.

In the previous post on this topic, Foundational Divides, I introduced the concept of an individual worldview as being represented by a single point in a very high dimensional “worldview space”:

Let’s assume that at a given point in time each individual has a set of values and beliefs, whether implicit or explicit, which provide them with a general framework for understanding and analyzing specific problems and deciding what the “right” thing to do is (i.e. what is their position on a relevant issue). We’ll call this finite, but very large, set of values and beliefs their individual worldview, and assume that if we knew all the relevant factors that construct their worldview at a given point in time, we’d be able to perfectly predict their position on any specific issue.
We can think of each relevant value or belief as identifying a spectrum along which individuals vary (in terms of how strongly they agree with or oppose the value or belief). Collectively, these value/belief spectrums define a high-dimensional space in which each relevant spectrum represents a single dimension, or vector within the space, and in which each actual or potential worldview is defined by a single point. Individuals who have very similar worldviews will be described by points which are very close to each other in this high dimensional space, while those who have very different worldviews will be described by points which are far apart.

The total “worldview space” could in theory be reasonably well approximated by a much lower dimensional model. Such a model would necessarily exclude important details, but could nevertheless be useful for communication and prediction. We can draw an analogy with the Big 5 personality factors, which represent a low dimensional model of the “personality space”, the origins and supporting assumptions of which I described in detail in Mapping the Mind. While having someone’s Big 5 scores is very unlikely to help you understand them better if you already know them well, it likely would tell you something relevant about someone you just met, and it provides a useful albeit incomplete framework for discussing and studying personality.

There are various models which aim to do this for worldviews too, after all we humans have a deep urge to categorize ourselves, especially into binaries—just think about how many versions of “there are two kinds of people in this world….” you’ve heard before. And in Foundational Divides I summarized three potential contenders for a primary or dominant worldview axis: progressivism vs. conservatism, the constrained vs. unconstrained vision (quite strongly related to progressivism vs. conservatism), and pomo vs. rationalist epistemology (which currently divides both the political left and the political right to the endless confusion of poor Sam Harris).

The structure of a worldview

Each of these provides an intelligible axis along which people differ, and which can be argued to be the primary cause of other, largely derivative, beliefs or opinions. However, the idea that there’s a core, fundamental axis suggests a specific model of how worldviews are formed and organized. One which resembles a decision tree, where your position on the fundamental axis constrains your range of possible beliefs, values, and views on a whole host of seemingly unrelated issues2. While there will be significant ideological differences among people on the same main branch of the tree, the set of options on which they’ll disagree is still bounded by what’s available given their chosen branch, by what’s consistent with the deeper shared value.

This underlying worldview structure is core to the argument made in Against The Popcorn Theory of Cultural Drift, in which

Rick

pushes back on what he sees as Hanson’s overly pessimistic view of our situation (which we discussed in depth in a recent Moral Mayhem episode). Rick argues that while many of the issues which Hanson notes as examples of drift are concerning, they aren’t at the core of the worldview which has made our culture so successful, and protecting and defending that core is the high leverage way to guard against a truly maladaptive (and Rick would argue, necessarily revolutionary) shift in values. Summarizing Robin’s argument that low selection pressures lead to drift and what he thinks it misses Rick says:

The memetic fitness landscape in which these new values thrive is decoupling from the old vicissitudes of hard work, discovery, improvement, and fecundity. The price of neglecting those vicissitudes used to be that our men got killed and our women got kidnapped by new husbands. But with the monoculture, there's no price for neglecting the old ways. So we drift. [...]
What’s going wrong here, I submit, is what often goes wrong with such arguments—the assumption of zero correlation. We don't need to defy probability theory, we only need some kind of correlation pattern among our values. Maybe “our values” is not just a grab-bag of isolated objects that all face cultural selection independently. Maybe there is some kind of hierarchical structure they're all a part of. Not a perfectly rigid structure, but one that connects them and makes them push and pull on each other when outside stresses get applied.

The core which Rick sees as in need of defense can be more or less summarized as “enlightenment rationality” as opposed to postmodernism (on the left and the right), as we touched on last time. But whatever you think the fundamental axis is given such a hierarchical structure, you’d need to make the argument in its favor on the basis that it explains more salient and relevant differences than other contenders.

Revisiting contenders for fundamental axes

For instance, if you think objective truth is out of reach, that truth claims, and claims to authority or expertise in general, simply reflect the latest winner of a power struggle, and that there’s no objective way to compare these claims on their merits, you’re going to evaluate evidence and arguments very differently from someone who thinks we can design processes which, while imperfect, at least approximate or bring us closer to truth, and that we can use reason to evaluate these processes and their results. When these sorts of differences aren’t made explicit, it can lead to very frustrating conversations, something I touched on in Has the right gone woke?

[I]f you believe that systemic forces bias all available empirical evidence in favor of certain preferred conclusions, the ones which benefit the powerful, why would you trust such evidence to support or disprove your claims? Under those circumstances anyone with power and anything which legitimates the position of the powerful is suspect. [...]
I’ve experienced this maddening dynamic in real life conversation. As you unearth various sources of evidence which fly in the face of the claim being made by your interlocutor, their never-ending cynicism invalidates each one until, exhausted and confused, you finally realize that there is nothing you could provide which they would see as legitimate evidence to falsify their claim. Because, they respond, of course that’s what the data would say because that’s what they, the powerful, want it to say!

From this perspective you can see how a postmodern conception of truth, paired with who you view as holding the power, will invalidate or validate a multitude of other claims. For instance, vaccine mandates and safety claims—outcomes of the scientific process, or convenient vectors to control and subordinate the population? Or the likely effects of housing deregulation—more positive sum transactions, and more housing for all income levels, or just another way for the fat cat developers to get rich at the expense of the common man? Etc., etc., etc.

Admittedly, many people who hold the beliefs I’m identifying with a postmodern conception of truth may feel that objective truth is possible in theory, it’s just that our current system is totally corrupt and therefore… revolution is necessary. And in general, to disagree with the postmodern conception of truth doesn’t imply that you think power is irrelevant. Just that you believe it’s at least possible to construct rational or adaptive processes for finding truth and designating authority, and that it’s possible to compare such systems and to evaluate truth claims in light of the processes that produced them. So while rationalism doesn’t preclude criticism of power structures, it does constrain the bounds and style of the criticism.

But there are other frameworks to consider. Thomas Sowell argues that individual worldviews are shaped by your vision of human nature and belief in the potential for radical progress. Those with a constrained vision see human nature and selfishness as a fixed trait that social and political systems must account for. In contrast, those with an unconstrained vision see such human flaws as correctible and believe injustice is not inevitable but the product of unjust systems. According to Sowell, these differing visions influence many seemingly unrelated beliefs and explain the alignment on so many issues within each political “side”. As he says in A Conflict of Visions:

One of the curious things about political beliefs is how often the same people line up on opposite sides of different issues. The issues themselves may have no intrinsic connection with each other. They may range from military spending to drug laws to monetary policy to education, Yet the same familiar faces can be found glaring at each other from opposite sides of the political fence, again and again. [...] A closer look at the arguments on both sides often shows that they are reasoning from fundamentally different premises. These different premises—often implicit—are what provide the consistency behind the repeated opposition of individuals and groups on numerous unrelated issues.

Sowell contends that those with the constrained vs. unconstrained vision hold entirely different conceptions of morally relevant concepts like “equality”, “freedom” and “justice”. Is “freedom” simply “the absence of externally imposed impediments” or does it also imply the absence of “circumstantial limitations which reduce the range of choice”. And while anyone can observe the very significant differences in income and wealth among individuals, on its own that observation implies nothing under the constrained vision in which “equality” is seen as a “process characteristic” (everyone is treated equally), but is proof positive of injustice in the unconstrained vision, where equality is seen as a “result characteristic” (everyone achieves equal outcomes)3.

Left vs. Right identifies fundamental differences but with strategically grafted on values and beliefs specific to a particular context

So the observed consistency of views among individuals who identify with a particular “side” can often be explained by their holding some more primary vision or epistemological framework in common. And it’s this more fundamental commonality which causes them to agree on these seemingly unrelated but actually largely dependent beliefs rather than this simply being evidence of social pressure to conform with an ingroup (although this likely plays a role as well). Still, as I mentioned in the last piece, left vs. right isn’t easily explained by means of a single axis:

While there are core values and themes which are consistently associated with each side, precisely what these terms reference shifts over time. Rather than identifying a clearly interpretable axis of difference, what counts as left vs. right is in part determined by what will separate individuals into two broad (and typically similarly large) clusters, such that most of the population can be roughly identified with one or the other.

I see the left-right binary as determined to some degree by how people differ in terms of a consistent ideological vision, but with the additional influence of context-dependent strategic alliances, such as the one between libertarians and religious conservatives on the American political right. And these strategic alliances mean that the baskets of beliefs we identify with the left vs. right are not all explainable as the outcome of a shared vision or a shared belief about truth or a shared understanding of the potential for progress.

The analogy with personality continues…

This multi-level understanding of worldviews, where deeper levels of core values are causally related to other observed positions, beliefs and behaviors strongly resembles models of personality. Quoting myself from Mapping the Mind:

[T]here are various theories related to the general structure of this personality system, the most well-known of which is probably the Freudian model. Freud emphasized the interplay between forces that shape personality, dividing it into three components: the id, ego, and superego, which represent unconscious drives, conscious awareness, and internalized social and cultural values, respectively.

This multi-level structure, whether in personalities or in worldviews, could complicate a data-driven approach to measuring the variance of relevance. Similar surface-level personality characteristics (or beliefs or values) could theoretically be motivated by different core drives (or fundamental visions or epistemological frameworks) which have been processed differently as a result of how they combine with intermediate traits (or facts or experiences) specific to that individual.

For example, Is someone a libertarian because they believe that individual freedom is a fundamental human right that must be respected unless there’s a very good reason not to do so, or are they a libertarian because they think the incentives created in such a system, in light of our imperfect and incalcitrant human nature, lead to more prosocial outcomes than do those implicit in other systems?

As it applies to personality, Raymond Cattell, whose early factor-analytic work laid the foundations for the eventual development of the Big Five model, noted the potential difficulties with teasing out the core, stable personality factors that could arise given such a structure. As he explains below, it’s possible that the same core drive could nevertheless manifest in very different outward expressions4:

The basic generalization we wish to stress here is that, among individuals possessed of equal endowment in a particular drive, different manifestations will vary inversely and be negatively rather than positively correlated. The situation may be explored more fully by means of Fig. I, in which we take a minimum population of two persons, possessing differing endowments in the basic erg (in this case sex drive) and differing amounts of investment in different manifestations.

And he includes the below diagram intended to illustrate the multilevel structure by which the outward personality expressions can be traced the to core “ergs”, similar to the decision tree structure I discussed above with respect to worldviews.

Cattell ultimately defined personality as “that which permits a prediction of what a person will do in a given situation”5, which limits the concept to things which can be observed, regardless of what the underlying structure looks like. But he expected that a properly defined factor analysis could indirectly detect the strengths of these “deeper, non-overt levels of dynamic integration” just as a “plumber could deduce the volume of water running through street conduits merely from observing the faucets in use in many houses.”6 And he implies that understanding this deeper structure would improve your ability to predict observable behaviors.

What’s the equivalent of the lexical hypothesis?

I’ve introduced the idea of both personalities and worldviews as residing in a very high dimensional space, as well as the goal of finding a low dimensional model for each space. But how do you define such a space, a step that precedes the work of determining an efficient lower-dimensional model for it, in an unbiased way?

The key assumption underlying the research program which ultimately led to the Big 5 was the lexical hypothesis: basically that any important personality trait, anything which describes a salient source of difference between individuals and which is relevant to the concept of personality, would’ve been given a name in any well developed language. Expanding on this point in Mapping the Mind, I noted that:

Once again, a core assumption underpinning the lexical approach is that, since the concept of personality is very important to us, we can assume that the set of all personality-relevant adjectives in a well-developed language with a large vocabulary will cover most if not all of what we care about. Therefore, we can assume that the set of vectors defined by these adjective-antonym pairs forms a “spanning set” for the space of personality. In other words, they will ‘cover’ (almost) all of the high number of dimensions which in aggregate define the personality space, and they will certainly cover the dimensions of personality which are most relevant or salient to us.
The assumption that our language spans the personality space is incredibly important since, if it holds, it provides us with an unbiased starting point from which to search for the most explanatory independent dimensions of personality. If a researcher was tasked with creating a personality questionnaire from scratch or with listing a set of relevant dimensions along which individuals could be ranked, it would be impossible to avoid baking in their personal biases. The questions they’d choose to ask or the dimensions they’d list would reflect their existing beliefs about the core drivers of personality, and so the results of any related analysis would also reflect those beliefs. (Previewing the next piece… if we were to approach the question “what is a worldview?” we’d want to think through what the equivalent of the lexical hypothesis for this application would be.)

So… what could potentially play the analogous role for worldviews that the lexical hypothesis played in personality research? I don’t know. Maybe you can examine the history of ideas, philosophy, the scientific method etc. But unlike with personality, where most of us have well developed vocabularies for describing the inner and outer characteristics of ourselves and others, and an intuitive understanding of the relevant ways in which people differ, most people’s worldviews and those of their interlocutors are never articulated. And so it’s not nearly as clear that the set of “worldview relevant” or “philosophically relevant” vocabulary or ideas would span the entire “worldview space”, nor is it clear that individuals could directly judge where people they know, even those they know very well, fall along the relevant axes.

And, depending on what the context of interest is, we might not have access to people who differ on the most meaningful axes. Particular cultures and societies are likely defined by only a narrow slice of worldview space which everyone in their population occupies. As Hanson has said, we tend not to tolerate much intra-cultural diversity on sacred values. Of course this is an issue with personality too, environmental and social factors affect the expression of your constitutional traits, but worldviews feel much more culturally determined than do personalities, and the total range of possible worldviews feels less constrained.

A worldview is some sort of complex, multi-layered construct that shapes and constrains our beliefs, values, and positions on a wide range of issues. And the parallels with personality research suggest some directions for further exploration, but also various potential pitfalls. Just as the Big Five model offers a low-dimensional framework for understanding personality, identifying analogous principles for worldviews could illuminate the foundational differences in our perspectives.

I don’t have answers, but I do find this framing helpful, and I hope some of you did too! If so, please consider subscribing or upgrading to paid—and if you’re interested in talking more about Sowell’s constrained vs. unconstrained vision please RSVP here for the upcoming book club event (for paid subscribers only) on Monday November 25 from 6:30 to 8:00 pm EST. Hope to see you there!

All Cats Are Female

Discussion about this post