Moving Beyond System 1 and 2

Can you think without thinking?

In Unforgiving Places: The Unexpected Origins of American Gun Violence—a book featured prominently in the Behavior series, which I’m still finishing¹—Jens Ludwig recounts a seminar given by the economist Richard Thaler on moral perception. During the talk, Thaler described a small experiment that, to Ludwig, revealed something big about human thought.

In the experiment, participants were presented with slightly different versions of the same scenario. One group was told that Gap—the clothing retailer—gave 50 percent of its “Red” campaign profits to charity. This group rated the company favorably. A second group was told the same thing, but in addition, that Gap kept the other 50 percent; this group’s impressions of the company plummeted. A third group was told both facts, plus that Gap could have kept all its profits; their ratings matched those of the first group.

Ludwig reads these results as evidence of what Daniel Kahneman calls System 1 thinking—the quick, automatic, unreflective side of the mind. “People are not taking the briefest extra moment to reflect,” Ludwig writes. “People are not thinking.”² If they were, he argues, they’d have realized that no new information had actually been introduced.³

Of course, Ludwig’s reason for mentioning the experiment is to make a broader point about gun violence:

Under conventional wisdom, violence interrupted is merely violence delayed. But behavioral economics gives us a way to understand why violence interrupted can often be violence prevented: It is so often due to System 1 motivations that can be fleeting in the face of time or with a bit more System 2 reflection.⁴

In other words, when we’re thinking fast instead of slow, we’re more vulnerable to making mistakes. Ludwig interprets the effects found in the Gap study as an example of a fast-thinking error—specifically, our tendency to ignore what isn’t explicitly stated, known as omission neglect. “That’s why,” he writes, “a respondent who’s told that the Gap gives away 50 percent of their profits has such different feelings…than does someone told that the Gap gives away 50 percent of their profits but keeps the other 50 percent.”⁵ In the first version, keeping 50% of the profits is implied but unstated, and therefore easier for someone operating in System 1 to ignore.

Ludwig’s prescription is for more System 2 thinking—more conscious, deliberate thought. If we could only slow down and reflect longer, he suggests, we’d make better moral, social, and even violent-crime decisions. Hence his enthusiasm for programs such as Becoming A Man, which teach teenagers to pause, reflect, and reframe their automatic responses before they escalate.

The Laziness of the Laziness Answer

If System 2 produces better decisions, why don’t people use it more? To me, the usual answer—that humans are mentally lazy—doesn’t cut the mustard. It’s undoubtedly true, as Kahneman says, that “laziness is built deep into our nature.”⁶ In other words, organisms are designed to conserve energy when and where they can. But…isn’t making better decisions exactly the kind of thing evolution would want us to spend energy on?

Let’s return to the study itself. I doubt that participants were unaware Gap would keep the other 50 percent of its profits; that fact is heavily implied. So why didn’t participants factor it in? Because they weren’t prompted to. Far from acting as rational calculators of objective truth, participants were trying to read the social situation. In that task, an important consideration becomes: Why is the experimenter sharing this particular information?

When an experimenter highlights a fact that seems redundant (“Gap keeps the other 50 percent”), participants may treat its inclusion as meaningful— even if they can’t say why—because if it were pointless, why would the experimenter mention it? From this perspective, participants aren’t being lazy, rushed, or irrational but socially attuned. They are inferring significance from the framing itself. The mere presence of information becomes a cue, encouraging participants to use it.

Indeed, there is a robust literature on word problems showing that people often struggle with them because such problems violate conversational pragmatics—what linguists call Gricean norms. For example, if I ask whether you have a watch, I don’t care about the answer; I’m really asking for the time. An experiment could, in theory, exploit this gap between literal and implied meaning to show that people are lazy and ought to think harder. But would that be a fair conclusion?

The influence of these conversational norms may be particularly strong in matters of moral judgment, which the Gap study elicits by asking participants to judge the company. As Rob has argued at length, morality functions as a coordination mechanism: it helps people get on the right side of a conflict before that conflict occurs. I would argue that the desire to be on the right side of a moral judgment drove participants’ responses more than the desire to think through the problem like a good economist. As such, participants took the inclusion of information as a cue that they should use it, given that throughout evolutionary history, if an interlocutor mentioned something, it probably meant something.

Of course, this moral bandwagoning approach isn’t ideal for empirical accuracy, but it is well suited to the evolutionarily critical task of reading the room—of inferring the norms and motives at play. In other words, sometimes you have to be wrong to belong.

The Modular Turn

The modular view of the mind—the one favored by evolutionary psychologists—reframes these results. Rather than toggling between System 1 and 2, the mind is composed of many specialized systems shaped for survival and reproduction, with the active one depending on the situation. Different situations activate different modules, not different “levels” of thought.

People in the Gap experiment weren’t failing to think; they were thinking with the module that best fit the context.⁷ From this perspective, the fault lies less with the participants than with the researchers, who ended up measuring how people respond to conversational pragmatics and the dilemma of being on the right side of a moral judgment, rather than how they evaluate information in isolation.

From a modular perspective, our so-called “cognitive biases” are not bugs but reflections of design features operating under conditions different from those they evolved for. Specifically, they reflect two kinds of mismatch:

1. Experimental mismatch. Some distortions arise from lab conditions that activate the wrong module for the task at hand. The researcher (especially if they are an economist) wants an independent thinker who pursues objective truth; the experiment instead primes a tribal member attuned to belonging.

2. Environmental mismatch. Many biases stem from the modern world itself, which differs in crucial ways from the ancestral environments in which our decision-making heuristics evolved. Still, our “fast and frugal” heuristics perform far better than we assume. The recognition heuristic, for instance—our tendency to favor or trust what’s familiar—may look irrational in modern life (brand loyalty, for example) but is, on average, surprisingly accurate.

In both cases, the problem isn’t that humans are lazy or failing to think hard enough—it’s that researchers have a partial view of human psychology. The issue lies not with the heuristic, but with the scenario—experimental or modern—that happens to trigger it. Indeed, the consistent replication of these effects should be seen not as evidence of human irrationality, but of the remarkable reliability and context-sensitivity of the mind.

So let’s return one last time to the Gap study. The module activated in participants was likely one concerned with fitting in, aligning with others, not ending up on the wrong side. The information given to each group was read as a hint to use it, which explains why their judgments flipped with each addition. Participants were “wrong” only because the scenario was artificial. Again, inferring moral intention from redundant signals—those that echoed rather than added new information—would often have been adaptive throughout human history.

Most of this, by the way, happens unconsciously. Participants aren’t aware that they’re solving a coordination problem—that they’re trying to end up on the majority side of a potential conflict. And they’re unaware because it isn’t necessarily adaptive to be aware. One doesn’t need to understand how or why they wind up on the winning side—they only have to wind up there.

Let’s revisit Ludwig’s quote:

Under conventional wisdom, violence interrupted is merely violence delayed. But behavioral economics gives us a way to understand why violence interrupted can often be violence prevented: It is so often due to System 1 motivations that can be fleeting in the face of time or with a bit more System 2 reflection.

A modular view would put it differently. When violence interrupted becomes violence prevented, it’s usually not because a person finally engages System 2, but because they’re no longer in the same situation—and may never be again.

References

Ludwig, J. (2025). Unforgiving places: The unexpected origins of American gun violence. University of Chicago Press.

1. You know how it’s easier to work on pretty much anything other than the thing you should be working on? I wrote this essay while procrastinating on Part III of Behavior—a series which itself was a procrastination of finishing my Taxes of the Built Environment series, which itself was a procrastination of Contra CBT (still unfinished). Don’t worry, though, I’ll finish Behavior Part III soon…probably tomorrow.

2. p. 131, italics original

3. This way of thinking about thinking is widespread in economics. For some background, see Gigerenzer’s article, which opens this way: “Some years ago, at the Center for Advanced Study at Stanford, one of my economist colleagues concluded a discussion on cognitive illusions with the following dictum: ‘Look, either reasoning is rational or it’s psychological’.” The mistake here is that if decisions and judgments are not made by the brain—which thinks—then what else are they made by? Similarly, what other process—if not thinking—is occurring?

4. p. 158

5. p. 135

6. Thinking, Fast and Slow, p. 35.

7. Well, the module that would have best fit similar contexts over evolutionary history.