Rational Rationalization

Why it's good to believe your own bullshit

Here, I discuss an interesting new theory of rationalisation and the responses to it.

Rationalization is bad, right?

Rationalization has a bad reputation, and for good reason. We’ve all looked back on some past decision - ignoring red flags in a doomed relationship, doubling down on a failing project, skipping an exam study session because "grades don’t really matter"-and cringed. If only we’d seen through our own excuses sooner. If only we could go back and shake ourselves out of our self-serving delusions.

Instead, we invent convenient fictions to justify our bad decisions, both to ourselves and to others. Bombed a test? Well, you didn’t really care about it anyway. Failed to quit smoking? Maybe smoking isn’t that bad after all. Spent too much money on a pointless gadget? Oh, but this one will totally make you more productive. The problem isn’t just that we make excuses - it’s that those excuses keep us locked in self-destructive patterns. Rationalization can mean clinging to a failing research project long after the data has spoken, or a gambler convincing himself that just one more bet will turn everything around.

Rationalization doesn’t just preserve our personal delusions - it also props up larger societal ones. System Justification Theory, suggests that rather than confronting inequality, we often convince ourselves that things must be fair, because the alternative is too uncomfortable. One can just think "someone smarter than me would’ve fixed it if it were a real issue" and go about our day. Or, more cynically, maybe we’re just rationalizing our own inaction, why shake up the status quo when we’re doing just fine? (See Laurin & Jettinghoff, page 27.)

Its fairly easy to make people rationalize in the lab. Offer someone money to defend a viewpoint they don’t believe in, and they’re less likely to convince themselves it’s true because they can always fall back on "I was just saying that for the cash." But ask them to defend the same viewpoint without payment, and suddenly, their beliefs start to shift. It’s the same logic even pigeons appear to follow - they prefer food they previously had to exert effort to get, thinking “well if I had to work this hard to get it, it must be good right?” Presumably, their tiny bird brains aren’t consciously thinking, “If I suffered for this, it must be worth it”. This is probably for the best, if they were thinking, they’d almost certainly be plotting something sinister.

What’s particularly interesting is that we don’t just lie to others - we lie to ourselves. We don’t merely spin convenient fictions, we believe them. But why?

Psychologists have traditionally seen rationalization as a defence mechanism. We do it to protect ourselves from guilt and shame, to maintain a coherent sense of self, and to resolve the discomfort of cognitive dissonance when our actions and beliefs don’t align. Another take is that rationalization evolved as a social survival tool. If reason itself developed to help us win arguments, then believing our own bullshit, it is argued, makes us better at convincing others. After all, the best salesman is the one who genuinely believes in his own product.

Right?

These explanations have dominated psychology for decades. But more recently, Fiery Cushman, a moral psychology professor at Harvard, has proposed a completely different take: rationalization is good, actually.

Why am I writing about this? First, because reading Cushman’s work was one of the reasons I became a moral psychologist in the first place (though this particular theory isn’t about morality per se). Second, I’m currently working on a paper about Explainable AI, essentially, how we can get AI systems to justify their decisions in ways that make sense to humans. And since understanding how we explain ourselves is crucial for designing explainable machines, Cushman’s ideas felt like they might be useful to explore.

Also, the argument is interesting in its own right. You can check out the source, if you’re were keen to read the academic paper, which includes responses from others in the field (a wonderful feature of Behaviour and Brain Sciences - the journal where it was published).

How Is This Theory Different from the Others?

Let’s go back to something I said earlier:

We invent fictions to explain away our bad decisions.

Something is amiss with this description. It makes the assumption that when we rationalise our actions, its only when our decisions are necessarily bad. But according to Cushman, this view is, at best, incomplete. It assumes that rationalization is only ever about covering up one’s mistakes.

Instead, he argues that while our actions don’t always stem from rational deliberation, those actions may still encode useful information, information we often don’t necessarily have direct access to. And, crucially, rationalization might be one way to extract that hidden information. If that sounds confusing, don’t worry, let’s break it down by looking at the different ways human behaviour originates.

Four Origins of Human Behaviour

Consider these four broad categories of behaviour:

1. Instinct: Hard wired behaviours, encoded in our genes

2. Habits: Actions that were previously rewarded in some situations, and so reoccur in similar situations

3. Norm Conformity: Adherence to explicit or implicit social norms without necessarily having an understanding of why there should be such a norm.

4. Rational Action: Actions that are intended to fulfil some desired goal given one’s beliefs

Rationalization, in simple terms, is the tendency to treat every action as if it belongs in the last category - that we acted based on conscious, reasoned decision-making even when that’s not actually the case. Let’s take a closer look at each category to see where this assumption goes wrong.

1. Instinct

Some behaviours are hardwired by evolution. We don’t choose them in any meaningful sense; we just do them because they’ve been adaptive in the past. Cushman uses the example of infants instinctively pulling back from a steep drop:

Suppose that an infant crawls to high point and then pulls back from the edge by instinct (Gibson & Walk 1960). This action does not reflect a belief that heights are dangerous, or the desire to avoid falling; rather, the infant pulls back from the edge by instinct alone (Gendler 2008)

This specific example is a bit controversial as some argue that infants learn to fear heights rather than being born with that fear. But other instinctual behaviours are harder to dispute. Think about how you recoil from the smell of spoiled food or yank your hand away from a hot stove. You don’t sit there deliberating about bacteria or burn treatment - you just react. These reflexes don’t involve rational deliberation; they’re built-in safety mechanisms.

Interestingly, some researchers argue that moral judgments can operate in a similar way. Studies in moral dumbfounding show that people instinctively judge certain harmless-but-taboo actions (e.g., consensual sibling kissing in a one-time, consequence-free scenario) as morally wrong, despite being unable to articulate why. Their disgust reaction feels like a reason, even if it isn’t one in the usual sense.

2. Habits

While instincts are shaped by biological evolution, habits are shaped by reinforcement learning, our environment rewarding and punishing certain behaviours until mostly advantageous behaviours become second nature, and disadvantageous behaviours are extinguished.

Cushman gives the example of someone turning on a light when they enter a room:

A person might habitually flip the lights on when they walk into a room because it is typically useful (i.e., rewarding). Each time that the behavior is performed and rewarded, habit learning strengthens the mapping from stimulus to response. As a result, executing habitual action requires little cognitive effort. A person does not have to consider desires (“I need light”) and beliefs (“switches cause light”) to derive the value of performing an action; rather, the behavior is habitized based on its value in the past.

Even when you know the bulb has gone, you may still habitually switch the light on.

You can also see this with rats in a lab. Suppose the rat learns to press a button to receive food. Eventually, it might start pressing the button even when it’s not hungry or even when the button stops dispensing food altogether. Why? Because the act of pressing has been rewarded so many times that it has become valuable in and of itself.

This is an example of what is called model-free reinforcement learning; the rat is not learning a model of the world (i.e. learning what the button does), it is just assigning some positive value to the act of pressing the button. This kind of learning is characteristically inflexible. Unlike rational planning, where we adjust our behaviour based on new information, habitual actions persist even when circumstances change. The rat keeps pressing the button because, at some point, pressing the button was a good idea and that history outweighs the current reality.

3. Norm Conformity

A large portion of human behaviour isn’t driven by instinct or habit, but by social norms. Unlike instincts (which are innate) and habits (which are shaped by direct experience), norms are learned by observing and interacting with other people.

Humans are uniquely good at cultural learning, and some argue that this ability, and not raw intelligence, that is the real key to our species’ success (see also here). We adopt behaviours simply because they are normal, and we distrust those who fail to conform.

Norm-following is particularly relevant to rationalization because when we try to make sense of social behaviour, we often engage in third-person rationalization, that is, we explain other people’s actions in terms of their assumed beliefs and desires, even when their behaviour was shaped by unconscious social pressures.

For example, if we see someone following a cultural norm, we might assume they believe in that norm or value the behaviour in question. But in reality, they might just be going along with it because because that’s how humans tend to relate to norms.

4. Rational Action

Now we get to the category that we often think explains most of our behaviour: rational action. A rational action, in this sense, is one that results from explicit reasoning. It’s an attempt to satisfy some set of preferences based on some set of beliefs about the world.

Importantly, this definition is descriptive, not prescriptive. It doesn’t mean the action is necessarily a good one, or that the reasoning behind it is correct. It just means that the behavior resulted from by a goal-directed thought process.

For example, imagine someone meticulously planning a road trip. They look up the best routes, calculate gas costs, check the weather forecast, and so on. Their behavior is rational because it stems from deliberate reasoning about how to best achieve their goals. And this would be true even if the route they chose ended taking them the wrong way (e.g. because they were holding the map upside down).

But, as we’ve just seen, not all behaviors originate this way. And this is where rationalization comes in.

Explaining ourselves

We have a strong preference for explaining behavior, our own and others’, in terms of rational action. We like stories where people do things for reasons.

In fact, we like these stories so much that we invent them. This is the essence of rationalization: we take behaviors that weren’t actually the product of rational deliberation and retroactively construct explanations as if they were.

Even when someone insists, I didn’t think that! I don’t want that!, we’re quick to say, Ah, but deep down, you do. (This is how Freudian psychoanalysis gets away with much of its hand-waving.)

Rationalization works similarly. When we act we often don’t know why we did what we did. But when asked, we feel compelled to provide an explanation, and the kinds of explanations we give tend to take a specific form. We generate a post-hoc justification that makes it seem like the behaviour was rationally chosen. And then, crucially, we come to believe that explanation ourselves. Our brains are basically running PR campaigns, producing reasons for actions that were never reasoned into to begin with. But these just-so stories are often very poor explanations for human behaviour.

This is where Cushman’s theory takes a turn from the usual critique of rationalization. Traditionally, psychologists have viewed this process as a form of self-deception - a bug in the system. But what if it’s actually a feature? What if, rather than just making excuses, rationalization helps us extract real information from our own behaviour; information that our conscious mind didn’t originally have access to? Reasons there were perhaps never represented in the mind to begin with.

Free-Floating Rationales

The late philosopher Daniel Dennett, who we sadly lost just last year, was deeply fascinated by the explanatory power of evolution.

Animals with relatively simple brains often exhibit behaviours that seem highly sophisticated. Dennett sought to dispel the illusion that these behaviours necessarily imply complex cognition. Some actions may make an animal appear highly intelligent, but that doesn’t mean it possesses deep, deliberate understanding.

A classic example is the distraction display of ground-nesting birds. When a predator approaches, the bird performs an exaggerated "broken-wing" act, luring the threat away from its nest. At first glance, this seems to require an advanced understanding of deception, the bird appears to have a belief about the predator’s intentions and a desire to protect its offspring. But as Dennett explains:

This seems to be deception on the bird’s part, and of course it is commonly called just that. Its point is to fool the predator. Now if the behavior is really deceptive, if the bird is a real deceiver, then it must have a highly sophisticated representation of the situation [...] It is unlikely in the extreme that any feathered “deceiver” is an intentional system of this intelligence. A more realistic soliloquy for any bird would probably be more along the lines of: “Here comes a predator; all of a sudden I feel this tremendous urge to do that silly broken-wing dance. I wonder why?

This is a prime example of how we are tempted to rationalize behaviors - to explain them in terms of beliefs and desires even when none exist. The bird does not intend to deceive; it simply follows an evolved instinct. Dennett called this phenomenon competence without comprehension - the bird’s actions are functionally effective, even though it lacks an understanding of why they work.

But while the bird itself does not possess the rationale for its actions, that rationale still exists. It is not an illusion, it is encoded by evolution. Dennett called these free-floating rationales - reasons that exist independently of any agent’s conscious thought. The evolutionary process “discovered” a solution (the distraction display) that works, even though the bird does not comprehend it.

Rationalization Is Good, Actually

Why, then, would rationalization be beneficial?

Cushman argues that traditional accounts of rationalization overlook something crucial: non-rational behaviors (instinctual, habitual, or norm-conforming actions) are not necessarily mistakes. In fact, they are often quite adaptive. Instincts exist because evolution had good reason to select them. Habits persist because past rewards can reliably indicate future rewards. Social norms culturally evolve because they tend to foster cooperation and survival.

The problem is that, when we act on instinct, habit, or norms, we are often unaware of the underlying reasons for our behaviour. But through rationalization, by generating post-hoc explanations, we can uncover and integrate those reasons into our conscious decision-making (a process Cushman calls "Representational Exchange"). This allows us to transition from blindly following instincts, habits, and norms to deliberately applying their wisdom in future planning.

Returning to our earlier example, imagine if the bird could understand why its distraction display was effective. Instead of merely acting on instinct, it could apply the rationale flexibly in different situations. Cushman suggests that this is precisely what humans do: rationalization allows us to extract and integrate the knowledge encoded by biological evolution, reinforcement learning, and cultural evolution, improving our future reasoning and decision-making.

Take moral judgments, for example. We often form them quickly and intuitively, but we struggle to articulate why we hold them. This is one reason why moral philosophy has been debated for millennia, people often act on moral intuitions first, then try to construct justifications later. But as Cushman puts it:

There is something important to be learned about what one ought to do simply by observing the actions that people, including yourself, actually perform.

The "useful fictions" we generate, our post-hoc explanations for why we acted, are useful precisely because they allow us to refine our future reasoning. Rationalization, at its best, attaches free-floating rationales (the wisdom of evolution, experience, and culture) to our own conscious beliefs and desires, helping us make better decisions in the future. Where there was once competence without comprehension, rationalization enables actual comprehension.

For example, our innate aversion to rotting food and dangerous heights is deeply adaptive. Over time, rationalization transforms these automatic reactions into explicit, rational beliefs - "Spoiled food is dangerous; I should avoid it" or "Falling from heights can kill me". Though these beliefs did not cause the initial instinctual reaction, they ultimately align with reality and reinforce adaptive behavior.

Cushman offers another example:

Suppose that a person instinctively recoils from snakes. This instinct is adaptive because many snakes are venomous, but she happens to be unaware of this. Rationalising her instinct (i.e., attempting to explain her act of recoiling in terms of beliefs and desires), she might adopt the belief that snakes are dangerous. Similarly, she could rationalise her behaviour by adopting a general desire to be far from snakes, and this is a useful desire. In either case, the outcomes of her future reasoning are improved.

Rational Inference

Of course, just because we can construct justifications for our actions doesn’t mean they will always be correct. There is a risk that rationalization could lead us to adopt false beliefs. Returning to the above example, Cushman notes that someone:

might have concluded instead that snakes breathe fire, shoot crossbows, or dredge up hurtful memories of adolescence. Such beliefs would explain one’s instinctive recoiling, but they are false. Or she might have adopted the desire to avoid all animals or all things that are long and straight.

If rationalization frequently led to such false conclusions, it would be maladaptive. But Cushman argues that rationalization is constrained by rational inference, it is biased toward generating reasonable beliefs that are likely to be true. Instead of producing arbitrary justifications, it tends to operate within the boundaries of Bayesian reasoning, (something about AI and Zombies), using probabilistic inference to construct explanations that make sense given the available evidence. Specifically, rationalisation is said to integrate models of Bayesian inverse planning. However, the details of that fall beyond my intended scope for this post.

Conclusion

Rationalization, rather than being purely self-serving or delusional, is often a mechanism for discovering truth. It allows us to approximate rational explanations for behaviours that were not, in themselves, originally a result of rational processes.

I will end with a somewhat paraphrased quote from the conclusion of the paper:

Why did I do that? We ask ourselves this question often, but for certain actions, any answer to the question "why did I do that?" which invokes beliefs, desires, and reasons is at best a useful fiction. Whether or not we realise it, the question we are actually answering is: "What facts would have made that worth doing?"