Beyond Heuristics: When Escalation Becomes Rational (And Why It Still Fails)

The sunk cost fallacy is not a fallacy at all—it's a rational response to uncertainty that we've been taught to despise.

This inversion matters because it reframes one of behavioural economics' most durable findings. When Kahneman and Tversky first documented how people throw good money after bad, the interpretation was straightforward: humans are irrational creatures, slaves to their emotional attachment to prior decisions. The prescription followed naturally—eliminate the bias, make better choices. But this framing obscures something more interesting: escalation of commitment often emerges from a perfectly defensible decision logic, one that only becomes destructive when the underlying conditions change.

Consider a pharmaceutical company eighteen months into a drug trial. They've invested $40 million. The early signals are mixed—not catastrophic, but not compelling either. A rational actor, we're told, should ignore the sunk cost and evaluate the project on forward-looking metrics alone. But this advice assumes perfect information about the probability of eventual success. In reality, the company faces a different calculation: they've learned something about the drug's behaviour, the regulatory environment, and their own capabilities through that $40 million investment. That knowledge has value. Continuing might be rational precisely because they now possess information a fresh evaluator wouldn't have. The decision to escalate isn't irrational—it's a bet on asymmetric information.

The problem emerges when this logic persists beyond its rational shelf life.

Organisations develop what we might call commitment momentum. The initial decision to continue becomes self-reinforcing. Teams are hired around the project. Reputations become entangled with its success. Executives publicly champion it. The organisation's identity begins to fuse with the initiative. At this point, the decision to escalate is no longer primarily about forward-looking probabilities. It's about identity preservation, status maintenance, and the psychological cost of admitting error.

This is where Kahneman's insight becomes genuinely useful—not as a description of irrationality, but as a map of the psychological mechanisms that hijack rational decision-making. We don't escalate because we're stupid. We escalate because we're human, and humans are exquisitely sensitive to threats to their self-image and social standing.

The escalation trap has a particular sting in organisational contexts because it's often invisible to the people caught in it. A manager reviewing a struggling project doesn't experience themselves as defending their ego. They experience themselves as being appropriately cautious, as having valuable insider knowledge, as being prudent stewards of institutional learning. The narrative feels rational from the inside. The bias operates precisely through this sense of justified reasoning.

What makes this especially treacherous is that the conditions for rational escalation and irrational escalation are nearly identical in their early stages. Both involve incomplete information. Both involve sunk costs that shouldn't matter but psychologically do. Both involve teams that have developed expertise and commitment. The difference lies in whether the decision-maker can genuinely update their beliefs in response to new evidence, or whether the psychological machinery of commitment has already locked in.

The research on this is sobering. Once public commitment occurs—once a decision has been announced, defended, or resourced—the psychological barriers to reversal become formidable. Kahneman himself noted that people will often escalate commitment to prove that their original decision wasn't a mistake, even when the evidence clearly suggests it was. The escalation becomes a form of self-defence disguised as prudence.

The implication is uncomfortable: the heuristics and biases framework, for all its explanatory power, doesn't actually tell us how to escape these traps. Knowing that escalation bias exists doesn't inoculate us against it. Understanding the mechanism doesn't disable it. We need something else—a structural intervention that removes the psychological stakes from the decision, or creates genuine distance between the original decision-maker and the evaluation of whether to continue.

Until then, escalation will remain rational in theory and ruinous in practice.