Fitness Beats Truth

The The Interface Theory of Perception rests on a claim about evolution that sounds, at first, backwards: that natural selection does not favour creatures who see the world accurately. Donald Hoffman and his colleagues call it Fitness Beats Truth — and they argue it is not a hunch but a theorem.

Interface games

In a 2010 study Hoffman, with Justin Mark and Brian Marion, ran "interface games" — evolutionary simulations in which rival perceptual strategies compete. The world is a set of territories holding resources; an organism's fitness is the resources it actually secures. A truth strategy perceives the real quantities of everything; an interface strategy perceives only fitness-relevant categories — enough or not enough — far more cheaply. Across most of the parameter space the truth strategy is driven to extinction, because perceiving more of reality costs more time and energy while buying no advantage. Strikingly, the more complex the world, the worse truth does. The authors were careful not to overstate it: their simulations show selection can drive truth to extinction, not that it always must — but that alone overturns the textbook assumption that perception is "almost always" veridical.

The theorem

Hoffman's group then proved the general result. The Fitness-Beats-Truth theorem (Prakash, Stephens, Hoffman, Singh & Fields, 2021) shows that, for a broad class of worlds and fitness functions, the probability that a fitness-tuned perceptual strategy beats a truth-tuned one of equal complexity is at least (|X|−3)/(|X|−1), where |X| is the number of distinct perceptions — a figure that climbs toward certainty as perception grows richer. As Hoffman puts it, "as the organism gets more complex, the probability that veridical perceptions will escape extinction goes to zero."

The reason runs deeper than mere cost. In an evolutionary contest, he argues, truth is simply irrelevant: only the fitness payoff decides the outcome, and payoffs depend on the organism and its needs, not on the world alone. A slab of raw meat is life to a starving cheetah and worthless to a sated one — same reality, opposite payoff. "Fitness functions, not objective reality, are the coin of the realm."

The one escape

There is a single, crucial exception, and Hoffman states it plainly: "veridical perceptions escape extinction only if fitness varies monotonically with truth" — that is, only where seeing more truth reliably means more fitness. He argues such tidy worlds are vanishingly rare among real ecologies, where payoffs rise and fall with context. That monotonicity clause is the hinge of the whole debate, and worth keeping in view: the claim is not that truth never helps, but that nature rarely arranges for it to.

Sawing off the branch?

The sharpest challenge is a fair one: if evolution did not tune us for truth, why trust the evolutionary reasoning that led to this very conclusion? Hoffman's answer is that the theory is built from logic and mathematics rather than perception, and that the algorithmic core of Darwinism can be turned on Darwinism's own surplus assumptions — including the unexamined premise that space, time, and objects are objectively real. He likens it to Kurt Gödel using the axioms of arithmetic to prove their own incompleteness without thereby refuting himself, and to physicists using spacetime-based theories to conclude that spacetime is not fundamental: "our theories are ladders to new levels of understanding, and sometimes a new level of understanding leads us to kick away the very ladder that led to it." That implication for physics is taken up in Spacetime Is Doomed, and the reality Hoffman believes lies beneath the interface in Conscious Agent Theory.

Sources & talks

Hoffman makes the evolutionary case for a general audience:

Truth vs Reality: How we evolved to survive, not to see what's really there — Donald Hoffman (Big Think)YouTube

Primary sources: Mark, Marion & Hoffman, "Natural selection and veridical perceptions," Journal of Theoretical Biology 266(4):504–515 (2010) · Prakash, Stephens, Hoffman, Singh & Fields, "Fitness Beats Truth in the Evolution of Perception," Acta Biotheoretica 69(3):319–341 (2021) · and The Case Against Reality (W. W. Norton, 2019).