II. Norms and Perspectives

Social norms do a great deal to shape our interpersonal interactions. While social norms and descriptive norms only cover a subset of our actions, their shared nature make them a significant source of different cultural characteristics.^{Footnote 1} Social norms can govern the trivial, such as whether white can be worn before or after Memorial Day, or the much more serious, such as whether murder is the appropriate response to a perceived slight. Social norms can support or undermine the rule of law, and have a large role to play in promoting or inhibiting economic prosperity and political freedom.

Culture is tied in large part to social norms for two significant reasons: first, norms are shared across a population, and second, norms can vary between populations. Let’s consider each of these reasons in turn. If we follow Cristina Bicchieri’s definition of norms,^{Footnote 2} we know that social norms are behaviorally robust in part because they are Nash Equilibria. As Bicchieri argues, they are held in place because of our social expectations. In particular, individuals in a population P prefer to follow a behavioral rule R in a context C conditional on two kinds of social expectations: empirical expectations and normative expectations. Empirical expectations are one’s beliefs that most other people will follow R in C. Normative expectations are the beliefs that others think that the agent should follow R in C.^{Footnote 3} Because of the nature of this equilibrium, we know that they must be shared across the relevant population, and this must be known to all members of the population. It is in this way that we can say that norms are shared: everyone knows that everyone knows that people follow the norm. While norms may be equilibrium behavior, this does not mean that they are universal. As Bicchieri discusses in her definition of norms, social norms can transform a mixed motive game into a coordination game, as well as selecting an equilibrium of that coordination game.^{Footnote 4} Importantly, this implies that there are other equilibria that could be selected. Unless norms are strongly functionalist and always optimizing, we should suspect that norms will not be universal, but instead regionally variant.^{Footnote 5} That norms can be so different is part of what makes them fascinating areas of study, and clear contributors to culture. However, this variation leads us to some challenges in our understanding of norms. Is the emergence of norms entirely contingent and largely explained by random processes, or can we point to other elements of our social cognition that we can identify that help select and maintain social norms? Further, if we are interested in consciously changing norms, how do we improve our chances of success?

A standard sort of reply to the question of norm selection is that some equilibria serve as focal points. Thomas Schelling has compellingly argued that focal points often provide a coordinating device for equilibrium selection, even absent a material reason to choose.^{Footnote 6} For instance, if we are considering an account of fair division, a 50-50 split is common, in part because an even split is a focal point — it seems immediately salient to people, in a way that a 55-45 split doesn’t. A 50-50 split has the virtue of being symmetric, by offering each person the same amount. Each person is thus treated equally.^{Footnote 7} Focal points are salient and known to be salient — they pop out to us as obvious solutions. Suggesting something else requires an argument.

But as Sen has pointed out,^{Footnote 8} thinking about equality is a good deal more complicated than simply claiming that a 50-50 division is the way to achieve it. If we are concerned about work, we may think that the most salient criterion is the amount of effort each person put in to producing the surplus. If we are worried about natural merit, we may think the most salient criterion is the real contribution of each party, not just time put in. Finally, if our focus is need, we may find that the most salient criterion is the current distribution of goods, rather than anything that went into the production of the surplus. What we find is that there is not a single way of understanding the meaning of fair division — rather there are many possible ways of evaluating what “fair” means.

In each case, we have an example of a perspective informing us of the salience of different pieces of evidence for making our choice. Perspectives shape how we see the world by giving us an interpretive lens. This interpretive lens does the work of categorizing the things that we see. Not only that, it tells us what to pay attention to at all. The world is overflowing with information, and perspectives serve as our filters. Perspectives provide us with an account of what’s similar to what, what information is irrelevant, and what counts as most salient. They serve to organize our understanding of the world, and assist us in navigating it.^{Footnote 9}

Perspectives, therefore, mediate our experience of the world in a way that is not consciously obvious to us. Rather than getting the “raw” sense data of our sensory apparatus, and rationally, deliberatively considering how to best organize this information, perspectives filter and categorize for us. This is, of course, necessary: we would not be able to process every bit of information that we take in as if it were unique and worthy of our attention. We simply don’t have the cognitive resources to do so. So, instead, we create categories, sort situations by similarity to familiar ones, and ignore much of what goes on around us, to better focus on the things we care about.^{Footnote 10} This can cause us to miss information that we would want if we knew about it. Just as importantly, it can shape the information we do receive into existing categories. How something is presented to us can be as important as what is presented.

Because they are so epistemically foundational, we rarely notice being in the thrall of perspectives, but the evidence for them is substantial.^{Footnote 11} Not only are perspectives foundational, but there is clear evidence of their influence on both our reasoning and our choices.^{Footnote 12} While there is going to be interpersonal variation in what our default perspectives are, it appears that there are also at least some robust cross-cultural perspectival similarities.^{Footnote 13} While we each may apply our own perspectives in many situations, which perspectives we adopt appears to be heavily informed by our peers.

This is important for several reasons. Most importantly for this essay, it suggests that perspectives can play an important role in norm selection. If perspectives make certain evidence salient, then each perspective is going to direct attention toward those norms that are made more focal by the salient evidence.

As we are following the Bicchieri account of social norms, we see that they can be formally understood as equilibria of social expectations. In a population that broadly shares certain perspectives, and has some beliefs that most share such perspectives, this approach will immediately privilege the norms that are made focal as a result. While this does not guarantee that those privileged norms are ultimately settled on, it makes it much more likely that they are. This is primarily for two reasons: perspectives can help certain norms spread to new domains by offering an account of similarity of situation, and perspectives can filter evidence such that they increase the salience of a particular choice. We will explore each pathway.

In the first method, perspectives shape our understanding of the contexts in which particular norms apply. Norms can spread to new situations if the population largely agrees that the new situation is suitably similar to the old situation where a norm is in place.^{Footnote 14} For example, if we agree on what “fair” means in the context of sharing apples, it’s likely that we will find this an appropriate way to share oranges. It may not, however, extend to how to run a tax system. But determinations of similarity are going to be informed by perspectives. In a scientific context, this can be seen most easily by thinking about how models are deployed to describe widely disparate areas of the natural world. If we are using a model of a coupled oscillator as is common in physics, we may argue that fireflies are quite similar to synaptic connections and two grandfather clocks standing next to each other.^{Footnote 15} Each can be modeled as a coupled oscillator, and thus can be said to have a number of very similar features. But if we do not use a coupled oscillator model, these things seem to have very little to do with each other — a grandfather clock is not animate, fireflies are organisms, and pairs of neurons are small parts of much larger wholes. If we adopt a particular perspective, they are quite similar, but taking up another perspective makes them rather diverse. In this sense, shared perspective will result in shared sense of similarity between situations, and thus broader agreement about the proper scope of particular norms. In the work of Ross and Nisbett,^{Footnote 16} we see how the model of the Prisoner’s Dilemma can be taken to be an instance of communal cooperation when given the label of “The Community Game,” and can be made to be the domain of self-interested rationality when given the label of “The Wall Street Game.” The labels trigger a perspective, offering an account of what the situation is most like. This has a large influence over behavior, even though the material incentives remain the same across labels. The perspective helps to make salient existing norms that could be extended to the new context.

Of course, it is not only contrived lab experiments that can exhibit this kind of behavior change dependent on perspective. In terms of raw material incentives, it is a fairly safe bet to simply cheat on one’s taxes. Most governments rely on a voluntary compliance model of income taxation. Citizens are simply asked to pay their taxes. This is backed by the threat of audits and possible legal penalties, but in general there is far from sufficient enforcement to catch every tax cheat. So governments rely on citizens seeing paying their taxes as a duty that they ought to fulfill, and not an instance of maximizing their savings. That is, governments rely on people seeing the situation as “The Community Game” and not “The Wall Street Game.” This is likewise the case with all manner of civic behaviors, including filling out census information. In fact, the UK government Cabinet Office’s Behavioral Insights Team demonstrated that changing how tax payment was presented dramatically changed payment rates among delinquent taxpayers. Legal penalties did not move people, but highlighting that most of the delinquent taxpayer’s neighbors had already paid boosted tax payment rates among delinquents by 15 percent.^{Footnote 17} In the experiment, all that was changed was the wording of a letter, which reframed how the tax payment as to be understood, just as we saw with the Ross and Nisbett study. How we categorize things determines what information we respond to, and how we interpret that information.

On this approach, we see that perspectives in part help us select between different candidate sets of empirical expectations. Each of us has empirical expectations for communal cooperation, as well as empirical expectations for Wall Street. In general, we tend to see high rates of cooperation among neighbors, and even if we ourselves do not actively participate much in the stock market, we have beliefs about what behaviors prevail among traders. While norms are crucially supported by empirical and normative expectations, it can be an open question of which set of expectations we should be responsive to. Our social lives are rife with different social contexts, and we interact with many different groups of people. Perspectives help us detect which context we are in, and which population we are dealing with. It’s only then that we can have a stable set of social expectations on which to rely.

In this way, perspectives can allow pre-existing norms to spread to new contexts. We can take advantage of previous sets of social expectations to extend rules outward, provided that the new context appears to be suitably similar to the existing domain of the norm.

The second method relies on changing salience conditions. Sen’s example of different conceptions of fairness is a clear instance of this. As Sen discusses in “The Equality of What?” even the idea of equality is subject to our perspectives, as they make salient different things that we could be equalizing. Utilitarians think about utiles, whereas Rawlsians think about primary goods, and libertarians focus on freedoms. Each of these perspectives offers an account of equality, and only in rare circumstances would those accounts concur on a particular distribution. Each perspective privileges certain facts by making them most salient, and so ignores other evidence, treating it as less essential. A thoroughgoing Rawlsian would find it difficult to come to the utilitarian conclusion because her perspective would discount the evidence that the utilitarian relies on. Her Rawlsian perspective would naturally support notions of fair division that rely on primary goods as a basis of comparison.

This second method is importantly distinct from the first approach. In the first approach, we find a way for an already established norm to spread into new domains. In this second method, we find that perspectives can block other ways of reasoning from encroaching. The salient evidence that I attend to is what my perspective tells me is most salient. Other, potentially important, information can be filtered out if it is not relevant to a particular perspective.

What we find in this theoretical account is that social norms are not created in a vacuum. Instead, norms emerge in a social cognitive landscape that has already been shaped in large part by a combination of individual and shared perspectives. These perspectives make allied norms more likely to emerge, by creating the space for them. When allied norms do emerge, this creates the possibility for a mutually-reinforcing relationship between perspectives and norms. Each is validated by the other: perspectives are reinforced because norms allow individuals to reliably act on the categories made most salient by the perspectives, thus making them seem more natural. Norms are reinforced because perspectives narrow our conception of the possible. The actions prescribed by the norm can appear to be the only options, as alternatives are more difficult to conceive of. In this way, norms can take on an even more cognitive dimension through their interaction with perspectives, since even without any form of coercion, people take their choice sets to be narrowed down to the set of actions allowed by the norm, and made salient by the perspective. Those that fall outside of that set become more challenging to imagine, let alone act on.

This interplay between perspectives and norms can serve to lock in particular sets of behaviors, even if they are not good for a community, simply because there isn’t a cognitively available set of alternatives. Further, evidence that may help people be dissuaded from continuing the practice may not be appropriately salient, and thus may be discarded. This can make social norms even more robust than the Bicchieri account suggests, and thus even more resistant to change.

III. Empirical Support for the Perspectives Account

We can see robust empirical evidence for this claim particularly in the literature on aspirations and agency. Interventions designed to target perspectives, particularly around perceptions of women and girls, have demonstrated the importance of perspective shift for changes in behavior. Beaman et al examined the effects of a reservation system in Indian local elections.^{Footnote 18} India, through a constitutional change, required some local elections temporarily to mandate that some local officials be women. The reservation system was designed as a randomized trial, so some villages had no reservations, some had two terms of reservations, and some had one term of reservations. The effects of the two terms of reservations, in particular, were sizeable. Prior to reservations, residents who heard a political speech read in a female voice were more inclined to say the speech was not good when compared to hearing the same speech read by a male voice. This tendency substantially reduced after the reservation period. Women were also about 10 percent more likely to be elected to local political office after this reservation period ended. Perhaps most dramatically, girls were kept in school for longer periods of time, had access to more food, and had higher aspirations for their life prospects. The reservation system was an effort to normalize the idea of females with agency and political authority; the fact that it had spillover effects in girls’ education and nutrition suggests that the intervention did change the perspectives of the populations who had the intervention, which in turn weakened social norms that were anti-female in a number of different areas of life. By focusing on local elections, rather than, say, individual women, this allowed the intervention to operate on shared norms and perspectives. This not only increased men’s willingness to accept females in more prominent civic and economic roles, but it shifted parents’ aspirations for their daughters, resulting in greater investment. It also shifted the aspirations of women and girls themselves, as it introduced new possibilities. In the case of aspirations in particular, we can see the direct effect of narrow perspectives. If girls simply cannot conceive of having control of economic or political resources, they can’t possibly choose to seek that out. This is a sharp constraint on agency freedom. By directly challenging the perspective and creating opportunities for women to demonstrate that they could wield political power, interventions such as these can disrupt the foundations of harmful social norms that suppress the agency of women.

In an even more remarkable study,^{Footnote 19} experimenters set out to encourage a sense of autonomy through aspiration promotion in rural Ethiopia. A common issue in development is that many times, there are opportunities for high-return investments that are simply not taken. Sometimes this can be explained via loss aversion or material constraint, but often it is a sense of helplessness or nihilism: it is extremely common for the very poor to have a sense that they have no real control over their lives, and that their choices matter very little. With such attitudes, it is difficult for people to then invest in future gains, rather than smaller more immediate benefits, like increased leisure or other comforts. In sixty-four rural villages across Ethiopia, six households were shown four fifteen-minute videos that showed people in similar situations make effective choices that improved their well-being, with the aim of increasing aspiration and demonstrating that people who exhibit autonomy can in fact improve their station in life. Six other families were shown a soap opera with no particular aspirational message as a placebo treatment, and six others were shown nothing. In half of the villages, a larger number of families were shown the aspirational video in a big group. The experimenters measured aspirations before the video, right after the video, and six months later. They also measured (actual) savings and (hypothetical) demand for credit before the video and six months later. The study found that exposure to the aspirational videos led to a statistically significant increase in aspiration and sense of autonomy, as well as an increase in savings and demand for credit. This is all the more remarkable in that the intervention was only an hour of inspirational video shown once. The study is impressive precisely because it explicitly targets aspiration and autonomy alone, and nothing else. People’s beliefs and behavior changed in response to the discovery that their previous perspectives were holding them back. By exposing them to alternatives, it gave them more control over their lives, and helped to promote positive new behavioral outcomes.

In both of these studies, we see an expansion of people’s understanding of their choice sets, simply by making a wider array of choices more salient to them. This challenges dysfunctional perspectives, and offers new ones in their place. This change of perspective can lead to a change in social norms, as is most explicitly the case in the Indian example.

We should not understand this relationship between perspectives and norms to only hold in the development context. Even in privileged contexts, we can find this same kind of relationship. One of the most studied areas of this relationship in the West is with what is known as “stereotype threat.” Stereotype threat is an aspect of social identity theory in which people experience anxiety about conforming to negative stereotypes about their social group. This can often lead them to end up conforming to those stereotypes, as it tends to undermine performance. Taking a population of Asian-American women in a math class, researchers were able to increase or decrease the students’ performance on tests based on whether they had a pre-test exposure to a description of Asian-Americans as good at math, or women as not good at math. Control group students, for whom no identity was triggered, performed in the middle. Across conditions, performance varied by 11 percent.^{Footnote 20} Students conform to the expected level of performance for the salient social identity, even if they have both.^{Footnote 21} This is especially remarkable as this was a study conducted at Harvard University — the students in the study were all highly gifted and in an extremely privileged social position. And yet, the power of salience was still sufficient to change behavior.

In each of these three cases, we find strong empirical evidence that perspectives help determine which norms we take up — even at the most basic level of agency. The way we see the world around us, and how we fit into it, dramatically informs our effective choice sets. We cannot possibly exercise our own agency if we do not believe that we have choices. It is difficult to effectively plan our lives, or perform well in difficult circumstances, if we conceive of ourselves as part of a group destined to do poorly. On the other hand, once we see that we do have choices, and we are capable of performing, new behavioral outcomes are possible. Just as negative perspectives can hold us back, and help to reinforce unhealthy norms, positive perspectives that embrace our agency can support more healthy norms that embrace the exercise of agency.

IV. Perspectives and Change

Each of these three studies aims to reveal the causal structure between perspectives and norms by introducing an exogenous shock in the form of a perspective shift. Does this suggest that perspectives can only be shifted by some external force? I argue that it does not. External shocks are sufficient, but not necessary for a perspective shift. What is necessary is some variation in perspective that generates a behavior change that further generates some change in outcome that is salient to those who hold the dominant perspective. That is, even if the initial insight that led to a change is opaque to most at first, if they can observe the consequences of the change on their own terms, they can work backward from there. For instance, in a market context, we might imagine that an innovative firm could develop a process improvement that increases profits. This process improvement may have required a novel perspective to develop, but it can be evaluated on more general terms because there is wider agreement on what an improved outcome might look like. Market competition is such that there is strong motivation to adopt any business improvements that raise profits, and so these kinds of innovations will spread. Contrast this with a process improvement that does not improve profits, but instead reduces pollutants in a market where pollutants are unpriced and unregulated. There may have been a very real process improvement, but not one that will be salient for the broader population of firms. The outcome improvement is irrelevant on a perspective that is insensitive to pollution, and so the outcome improvement (and thus the process improvement brought on by a new perspective) would likely go ignored.

Just as with markets, social innovations brought on by novel perspectives would likely only gain purchase if they brought about benefits in outcomes as recognized by the dominant perspective. Fights for social equality can be easily seen through this lens: benefits of a more equal society are not transparent to perspectives that support gender, racial, sexual, or religious hierarchy and inequality. What’s worse, what appears to be a benefit in one perspective may appear to be a loss on another. Large development challenges, like child marriage, dowry systems, and female genital cutting appear to be instances of this. There isn’t just a question of adjustments of political power and authority with shifts away from these practices, but also a question of what is the natural social role for women and girls. Even if one had as a goal improving the well-being of girls in their society, without a perspective shift, it is unlikely that one could endorse a norm change away from dowries, child marriage, and other associated practices, in part because those could seem like worse outcomes. So while it is possible for perspectives to shift endogenously, there may be serious constraints on what kinds of new perspectives could have the potential to become dominant in a population, given its starting position.

As I have argued elsewhere,^{Footnote 22} changes in attitudes around gay marriage in the United States form a perspectival shift. As empathy toward gay people increased due to greater positive exposure of gays and more awareness of personal connections with gay people among the straight population, gay marriage has (rather rapidly) been re-categorized as being similar to straight marriage, rather than some different category. This made it easier for people to come to believe that equality meant support for gay marriage, where previous arguments contended that gay marriage was a break with equality, and a granting of “special rights.” Gay marriage has become broadly accepted in the United States not because novel legal arguments were created, or the Constitution changed, but instead because we changed how we understood what gay marriage is. It went from being a brand-new kind of behavior to just another instance of the pre-existing institution of marriage. Once our perspective changed, our behavior did as well.

Social norms are a crucial element of our social lives, as they shape a great deal of our behavior and our culture more broadly. But norms are not alone. They are supported or inhibited by underlying perspectives. This is important for two key reasons: first, it can help us understand how particular norms, rather than alternatives, emerged in particular societies. This insight can help enrich our understanding of the endogenous emergence of norms and culture. Equally important, however, is how this can inform any policy-directed efforts at changing existing norms or creating new ones. If this analysis is correct, it suggests that social norm creation that fails to take the role of perspectives into account may be more frail than we realize. Norms rely on the maintenance of empirical and normative expectations. Each of these expectations has a synchronic and diachronic element to it. It may be the case that we currently see everyone participating in a norm, but if that norm is out of step with our general view of the world, we may naturally discount that evidence as less salient than other factors as we look to the future.

The combination of perspectives and norms creates a powerful account of stasis. Social norms by themselves can be challenging to change, given their equilibrium nature. But social norms can conspire with perspectives to make it difficult for agents to recognize that there are even other options. Because perspectives are fundamentally subconscious epistemic filters, they can strip away choices that may be objectively present, but not epistemically accessible. Social norms and perspectives can thus reinforce each other, where compatible perspectives provide evidence for social expectations for the norm, and the norm highlights the salience of the values and behaviors made accessible by the perspective. This can make behaviors feel natural or inevitable, even if they are socially arbitrary, and potentially not in people’s interest. The most important aspect of this is that the conscious experience of following a social norm supported by a widely shared perspective may not feel like making a choice at all. There may be alternative actions available, but none that are consciously accessible.

V. Perspectives and Agency

Let’s now turn to what this can tell us about the nature of agency. On this account, agency is more nuanced than on the standard view. We have already seen that many choices are interdependent rather than independent, and perspectives narrow the scope of our perceived choice set. What’s most remarkable, however, is the degree to which perspectives and norms can conspire together to strip away or enhance the agency of an individual. This suggests at the very least that wide-ranging agency is not something that we can simply take for granted, neither in our descriptive work, nor in our normative work. Agency is an individual and social achievement — something to be nurtured and developed, rather than a fact of nature.

We have already seen several examples of perspectives and norms locking in place behaviors and attitudes that can reduce agency, whether it is across the board as in the Ethiopian case, or for women and girls in particular in the India and Harvard cases. This suggests first that agency is not an all-or-nothing phenomenon — it’s possible for a population to have mixed levels of agency in it. It also seems possible for agency to be less domain-general than our intuitions might suggest. Even in countries whose citizens have fairly extensive agency may find significant domains where that agency is simply not active. For example, Alesina, Giuliano, and Nunn^{Footnote 23} found that gender norms and the economic, political and social inequalities that flow from viewing males as the more dominant sex are strongly predicted by whether the society historically relied on plow-based agriculture.^{Footnote 24} In modern, non-agrarian societies, we have a series of behaviors, beliefs and attitudes held in place because generations ago men had an easier time using plows than women. On a standard rational actor model, this could only be explained by the idea that men and women continue to prefer these unequal roles. While it is possible that there are men and women who, in broad view of the structural inequalities present, continue to endorse those arrangements, it is much more plausible that this historical condition gave rise to a set of perspectives and norms that “locked in” behaviors that at one time may have made some material sense, but have lived past their usefulness. They have simply become so natural to us that they are not generally subject to conscious deliberation. Even when they are, the status quo is powerfully held in place: if one or a few people would prefer to change, they remain locked in place by the equilibrium nature of norms. Meanwhile, to convince others they need some evidence that is salient on the terms of the dominant perspective.

In this way, we may find that broadly shared social norms and perspectives combine to be enemies of agency. They do this in two ways. First, a perspective makes alternative options beyond the status quo hard to find; and second, even if a preferred alternative is discovered, the norm can lock status quo behavior in place. This is certainly the case in many areas. But we can also look to permissive norms rather than restrictive norms. Norms of tolerance, of mutual respect and recognition, and norms encouraging innovation and entrepreneurial activity (whether economic, artistic, or social) all can support, rather than inhibit agency. As Bavetta and Navarra^{Footnote 25} have argued, informal institutions can make agency freedom more possible. It is not simply the lack of norms that make agents more agential, but rather the presence of supporting norms that enables a wider choice set and the ability to freely choose.

One way to account for this, despite the evidence we have seen suggesting that norms and perspectives can conspire to reduce agency, is that these more permissive norms function in part by encouraging a proliferation of perspectives. Perspectives are most dangerous when everyone shares in the same one. This can create a social blindness: everyone fails to see options that are available to them, and does not recognize evidence in favor of alternatives as salient. When more perspectives are present in the population, however, we may find that we are closer to a situation in which new alternatives can be discovered and explored.^{Footnote 26}

On this account, agency is at least partly a function of the broader social environment. The fewer available perspectives in the population, the more likely agency is restricted. The ecosystem of social norms and perspectives shapes the frontier of individual agency across several domains — enabling a fuller sense of agency in some areas while restricting it in others. Perspectives serve both to limit options in the individual case, and provide a mechanism for expanding the option set in the social case. Restrictive social norms can hold behaviors in place even in the presence of a larger option set, but permissive norms, such as norms of tolerance, can preserve the broader option set.

This framework raises a challenge for economic modeling: How do we represent these phenomena, while keeping the clarity of the rational actor model? The rational actor model takes the option set to be a given, but if it is a function of the broader system of interactions, this idealization may be insufficient to capture the phenomenon. Perhaps the most important implication is normative: when evaluating the actions of others, we cannot merely take for granted that the person weighted all available options and acted according to his or her preferences, but instead, we have to determine what options that person took to be available, and whether social norms were in place that forced the person’s hand.

Perhaps most importantly, if we view agency to be something of value, then it is not merely a question of individual responsibility to act agentially, but a broader social and political responsibility to work toward adjusting our informal institutions to increase the domain of agency. In On Liberty, Mill recognized that coercion did not come solely from the formal apparatus of the state, but from the constraints of custom. In this essay, I have sought to illustrate the mechanisms by which the constraints of custom operate. These mechanisms suggest that no single individual can overcome them and fully exercise their own agency. Instead, we see that agency is a social achievement, made possible by an array of permissive norms, and a diverse range of perspectives. Denials of agency often come from broadly shared pairings of restrictive social norms and companion perspectives that conspire to eliminate choices and lock existing behaviors and attitudes in place. Individual agency is partly a function of a broader social ecosystem. So if we wish to promote individual agency, we need to look past individuals.