I started thinking about the psychological representation of modality—the way that we represent and reason over non-actual possibilities—sometime before graduate school (so, almost ten years ago). Over the course of the intervening years, Josh Knobe and I have been developing a theoretical framework for understanding this aspect of our minds, and now, the the paper where we lay out our approach is finally, officially, out!
While the paper is meant to be a lot of fun to read (see for yourself), I thought I’d also write a less academic blog post that leaves most of the details out and just gives a rough sense for the framework and why we think that understanding the psychological representation of modality is going to help solve a lot of problems throughout cognitive science.
I’ve tried to make this post pretty modular, so you can focus on the parts that you’re most interested in, and ignore the others. You’ll definitely want to start with the overview of our proposal about the psychological representation of modality, but the other sections that go on to apply this framework to different phenomena (how people make judgments about freedom and force, the problem of causal selection, the way kids think about possibility) can basically be understood separately.
The psychological representation of modality
How we talk about possibility
A good place to start in trying to understand how we represent non-actual possibilities is to consider how we talk about them. One intriguing fact is that across an enormous variety of languages, people often use the same term (e.g., English ‘can’) to make claims that seem to be about physics, as in (1), probability, as in (2), or morality, as in (3).
(1) Particles can’t go faster than the speed of light.
(2) One can’t complete an entire career in research without making a few mistakes.
(3) You can’t keep treating your sister that way – look at how upset she is!
What should we make of this? One possibility is that terms like ‘can’ are surprisingly polysemous across languages. However, linguists working in formal semantics have rejected this idea and instead argued that the meaning of all of these different kinds of claims can be captured within a single unifying framework (Kratzer, 2012). At a very informal level, you can think about the proposal as involving three key claims. The first is that statements involving modal auxiliaries (e.g., ‘can’, ‘must’, or ‘should’) are not concerned so much with the way that things actually are, but are really claims about (mostly non-actual) possibilities. The second is that people aren’t concerned with all possibilities, but rather, only a much smaller subset of them. And the third is that three things that can shape which subset of possibilities people consider are physics, probability, and morality.
To get a sense for this, consider how we might use this to make since of (1), for example. The idea is that this is a claim about not just the way that particles have moved thus far, but about all of the possible ways particles might move. Importantly though, we are not concerning ourselves with possibilities in which the laws of physics are violated, and clearly, once we exclude those kinds of possibilities, there are no cases of particles going faster than the speed of light.
We can also make sense of (3) in much the same way: It is a claim about not just what has already happened, but a whole range of possibilities involving you interacting with your sister. The possibilities we are not concerned with in this case, though, are ones in which the rules of morality are violated, and obviously, if we ignore those possibilities, there are no cases of you continuing to treat your sister that way.
Generalizing a bit then, the basic feature we’re pointing to is that when we talk about possibilities, we’re often concerned with possibilities that conform to what we know about physics, probability, and morality, and simply ignore possibilities that violate these factors.
How we think about possibility
With this basic idea in hand, our proposal is basically that a version of this picture is true not just of how we talk about possibility, but also how we think about possibility at a completely non-linguistic level.
You can think of our proposal as having basically two key claims:
(i) In any given context, people tend to represent not only the actual things that occur, but also non-actual possibilities.
(ii) Possibilities are not all treated equally: they tend not to be represented to the extent that they involve physical violations, morally bad actions, or highly improbable events.
Consider a concrete example. Imagine a person whose car breaks down on the way to the airport. She is now trying to figure out what to do next. At least in principle, she could consider an indefinitely large number of possibilities that involve her getting to the airport, including the following:
Hail a taxi
Call a friend
Steal someone else’s car
Convince the airport to delay the flight
Levitate and fly to the airport
Our claim though is that people don’t represent most of the possibilities that they (in principle) could. Rather, they just focus on the much smaller subset of possibilities that conform (among other things) to what they know about physics, probability, and morality. So, in our simple example, the possibilities the person would represent might look something more like this:
Hail a taxi
Call a friend
If you’re starting to get worried that we’ve spent all of this time saying something that’s just kind of intuitive and obvious, here’s when you can stop worrying!
The suggestion isn’t simply that people take different considerations into account whenever it makes sense to do so (perhaps by applying domain general reasoning principles). Rather, our claim is people will continue to take these same considerations (physics, probability, morality, etc.) into account even when it doesn’t make sense to do so. We’re suggesting that these constraints on the psychological representation of modality operate at a non-conscious level to inflexibly exclude possibilities from consideration. Thus, for example, we predict that people will tend to not to consider possibilities that would be morally bad even when they can easily see on reflection that there is no rational reason for doing this.
That’s the heart of the proposal; the only other real idea is that we think this non-conscious and relatively inflexible representation plays a fundamental role throughout higher-level cognition.
Putting modality to work
Alright, so now what we want to show how by positing this one kind of representation—the psychological representation of modality—we can start to explain a plethora of otherwise puzzling patterns of judgments across a diversity of areas of cognition.
We’ll focus in particular on three different cases where we think representations of non-actual possibilities play an important role and where one finds that violations of physics, probability, and morality have roughly the same kind of impact. Then, we’ll show the patterns are actually pretty easy to understand if you appeal to the psychological representation of modality.
So this is the part where you get to choose your own adventure. You can just keep reading, or if you’re only interested in one particular topic, you can skip directly to what we think the psychological representation of modality can do there. Here are three of the topics we consider (you can check out the others in the paper).
- How people make judgments about force and freedom
- The problem of causal selection
- The way kids think about possibility
Judgments of freedom
A central topic in both ancient and contemporary philosophy has been the distinction between cases in which a person acts freely and cases in which a person is instead forced to act. This distinction is at the foundation of many philosophical debates ranging from coercion to free will to political liberty (Aquinas 1952, Aristotle [340 BCE]2002, Berlin 1970).
While the theoretical discussion of these issues continues, experimental philosophers have sought to inform these debates by conducting studies that investigate how people ordinarily make these judgments. Unsurprisingly these studies have found that people are highly sensitive to what can physically occur. That is, they more judge that people are free to do something if they were physically capable of not doing it (see, e.g., Woolfolk et al. 2006). However, more recent research has also uncovered that morality may play a similar role: people are more inclined to judge that people did something freely if not doing it was morally good (Phillips & Knobe 2009, Young & Phillips 2011). In short, one finds that physical and moral considerations have strikingly similar effects on ordinary judgments of force and freedom.
The proposal we’ve argued for thus far is that people have a single representation of possibility (the psychological representation of modality) that is influenced by both physical and moral considerations. So now we want to put it to work by saying something about how judgments of freedom rely on this representation. Here’s the key idea:
- For people to conclude that an agent performed an action freely, they have to represent that it was possible for the agent to have not performed that action.
Critically, we’re not talking about people’s explicit or deliberative judgments of what would be ‘possible’ for an agent to do, but rather, whether this possibility is represented as part of their (non-conscious) psychological representation of modality.
With this piece, we now have a unified explanation for why physics and morality have the same impact on judgments of force: For people to judge that an action was done freely, they have to represent the possibility that the agent could have not done the action, and so if not doing the action would violate either physical laws or moral rules, then people will tend to not represent this possibility, and thus will not judge the person to have acted freely.
Since writing this paper, a bit more evidence for this theoretical proposal has been found. For example, measures of participants’ implicit representation of the possibility of actions (including immoral and improbable actions) are a much better predictor of judgments of force than measures of participants’ explicit understanding of whether these actions are possible (Phillips & Cushman, 2017).
Suppose that a forest fire was started by the simultaneous presence of oxygen, dry leaves, and a lit match. Which of these was the cause of the forest fire? Given just the physics of the scenario, it’s clear that the fire would not have started without all three elements. Typically though, people only think that the match caused the forest fire, or sometimes they might (depending on the circumstances) be inclined to regard the dry leaves as a cause, but almost no one would ever consider the oxygen to be a cause of the forest fire. The problem of causal selection is the problem of explaining why people privilege certain causal factors over others in cases like these.
A large number of researchers working in philosophy, computer science, law, and cognitive psychology have debated over the best way of formally model how causes are selected (e.g., Halpern & Hitchcock 2015, Hart & Honoré 1985, Hitchcock & Knobe 2009, Woodward 2006). One factor that has been noticed is that improbable events are often selected as the cause of the eventual outcome (see, e.g., Hart & Honoré 1985). However, probability is not the only factor which helps determine which causes people select. Another, perhaps surprising, factor is the moral status of the event. Specifically, people seem to pick out morally bad events as causes of eventual outcomes, even when the outcomes are neutral or good (Hitchcock & Knobe 2009). As a number of researchers have noted (e.g., Halpern & Hitchcock 2015), there is a notable similarity in these judgments between the effect of probabilistic considerations and the effect of moral considerations (Kominsky et al. 2015)).
To start, it’s worth pointing out that a number of existing accounts formally capture the effects of morality and probability in a unified way (e.g., Bello 2014, Blanchard & Schaffer 2017, Halpern & Hitchcock 2015, Icard et al. 2017, Knobe & Szabó 2013). While these accounts differ from each other in various ways, they also share one central feature: they all involve the consideration possibilities that differ from what actually happened (e.g., possibilities in which there is not a lit match, and thus no forest fire). At an abstract level, you can think about the core commitment across all of these accounts as something along the lines of:
- The judgment that some factor x was the cause of some outcome o involves reasoning over possibilities in which factor x differs in some way.
Our contribution is basically to combine this idea with our own proposal for how people actually represent possibilities. So, in the example of the forest fire, the thought is that the causal judgment in (4a) involves reasoning over possibilities picked out by (4b) below.
(4a) The lit match was the cause of the forest fire.
(4b) Possibilities in which the match was not lit.
Similarly, the causal judgment in (5a) involves reasoning over possibilities picked out by (5b).
(5a) The oxygen was the cause of the forest fire.
(5b) Possibilities in which there was no oxygen
Because the possibilities picked out by (4b) are highly probable, while the possibilities picked out by (5b) are highly improbable, we expect people to actually represent and reason over the possibilities in (4b) but not those in (5b). This difference is then precisely mirrored by causal judgments, which directly seems to reflect the fact that people are considering what would have happened if the match hadn’t been lit, but not what would have happened if there had not been oxygen in the forest.
The key thing about this approach is that it also allows us to explain why people tend to select immoral events as causes. The idea is that the causal judgment in (6a) will depend on the possibilities picked out by (6b).
(6a) The immoral action was the cause of the resulting problem.
(6b) Possibilities in which the immoral action was not done.
and the causal judgment in (7a) involves reasoning over the possibilities in (7b)
(7a) The morally good action was the cause of the resulting problem.
(7b) Possibilities in which the morally good action was not done.
So it’s not hard to explain why people are more inclined to agree more with claims of the form (6a) than claims of the form (7a) (Alicke 2000, 1992, Alicke et al. 2011, Knobe & Fraser 2008, Hitchcock & Knobe 2009, Roxborough & Cumby 2009, Kominsky et al. 2015). People tend to represent and reason over the possibilities picked out by (6b) but not those picked out by (7b). Thus, their causal judgments mirror this fact: they seem to be ignoring what would have happened if the morally good action hadn’t been done, but not what would have happened if the morally bad action hadn’t been done.
Development of thinking about possibilities
To adults, it seems obvious that certain things are possible (e.g., throwing one’s hat into the air), but that others are impossible (e.g., transforming one’s hat into a bottle of whiskey). Research in cognitive development has investigated how young children think about possibilities such as these, and how their understanding of them changes over the course of development.
Much like adults, it turns out that young children have no trouble judging that events that require violations of the laws of physics cannot actually happen (Levy et al. 1995). Yet, their understanding of possibility also differs from adults’ in remarkable ways. Children often explicitly judge that improbable events are impossible (Shtulman 2009, Shtulman & Carey 2007). Moreover, young children (3- to 5-year-olds) also judge that morally bad events can’t happen, are impossible, and even require magic to happen (Chernyak et al. 2013, Kalish 1998, Kushnir et al. 2015, Phillips & Bloom 2017).
In sum, we find a developmental pattern such that young children regard events involving physical violations, morally bad actions or statistically improbable occurrences to be impossible. Then, as children age, they increasingly tend to judge that only events involving physical violations are actually impossible. The question now is how to make sense of young children’s puzzling judgments and the changes that occur as they age.
While we expand on this idea in more detail in the paper our basic suggestion is pretty straightforward. Early in life, young children’s explicit judgments of possibility reflect the default way that adults implicitly reason about possibility. On this approach, what’s easy is explaining young children’s understanding of possibility. But, interestingly what becomes more difficult is explaining adults more reasoned judgments. That is, given that it is extraordinarily improbable to find an alligator under one’s bed, why do adults say that such things are possible?
To explain this kind of response, we’ll need to say something about the additional sophisticated capacity that adults develop. We suspect that one likely explanation is that adults have learned how to prevent certain factors from constraining the possibilities that are considering. For example, they may have realized that the question they are being asked is specifically about which events involve physical violations, and accordingly prevented other constrains (moral, statistical, and so on) from playing much of a role. In other words, it may be that modal reasoning defaults to taking into account a variety of different considerations (physical, moral, probabilistic), but adults have developed a capacity for a more sophisticated kind of reasoning that allows them to deviate from this natural default.
Once again, more recent research has provided a test of this hypothesis by comparing adults’ judgments of possibility when they reflectively deliberate to their judgments of possibility when they are forced to respond extremely quickly (and thus are less able to engage in any kind of sophisticated or effortful reasoning). When adults are unable to engage in sophisticated reasoning, their judgments of what is possible begin to strongly resemble those of young children (Phillips & Cushman 2017).
The paper itself
I think of this paper as a kind of birth certificate for the emerging body of work on the psychological representation of modality. It’s where we first wrote down the name and it gives some requisite acknowledgment of its (many) theoretical predecessors, including work by Angelika Krazter). I also think that there is an important way in which the paper poses more questions than it answers about the psychological representation of modality. All the same, I also think it outlines the nascent form of what some of those answers are going to look like in the next ten or so years.
The full paper, along with all of the relevant references, can be found here:
It’s also worth pointing out that Josh and I owe a huge debt of gratitude to a bunch of people who helped us think more carefully and deeply about this topic including (but definitely not limited to) Angelika Krazter), Fiery Cushman, Tobias Gerstenberg, Jonathan Kominsky, Matthew Mandelkern, Paul Boom, Jonathan Schaffer, Brent Strickland, Charles Kalish, Paul Harris, Laurie Santos, John Turri, Liane Young, David Rand, Justin Martin, Adam Morris and members of the Moral Psychology Research Group.
Chernyak, Nadia, Tamar Kushnir, Katherine M Sullivan & Qi Wang. 2013. A comparison of american and nepalese children’s concepts of freedom of choice and social constraint. Cognitive Science 37(7). 1343–1355.
Gerstenberg, T., Peterson, M. F., Goodman, N. D., Lagnado, D. A., & Tenenbaum, J. B. (2017). Eye-tracking causality. Psychological science, 28(12), 1731-1744.
Icard, Thomas, Jonathan Kominsky & Joshua Knobe. 2017. Normality and actual causal strength. Cognition 161. 80–93
Knobe, Joshua. 2010. Person as scientist, person as moralist. Behavioral and Brain Sciences 33(04). 315–329
Knobe, Joshua & Zoltán Gendler Szabó. 2013. Modals with a taste of the deontic. Semantics and Pragmatics 6(1). 1–42.
Kominsky, Jonathan F, Jonathan Phillips, Tobias Gerstenberg, David Lagnado & Joshua Knobe. 2015. Causal superseding. Cognition 137. 196–209.
Kratzer, A. (2012). Modals and conditionals: New and revised perspectives (Vol. 36). Oxford University Press.
Phillips, J., & Cushman, F. (2017). Morality constrains the default representation of what is possible. Proceedings of the National Academy of Sciences, 114(18), 4649-4654.
Phillips, J., Luguri, J. B., & Knobe, J. (2015). Unifying morality’s influence on non-moral judgments: The relevance of alternative possibilities. Cognition, 145, 30-42.
Shtulman, A., & Phillips, J. (2018). Differentiating “could” from “should”: Developmental changes in modal cognition. Journal of experimental child psychology, 165, 161-182.
Young, Liane & Jonathan Phillips. 2011. The paradox of moral focus. Cognition 119(2). 166–178.