Whenever I'm worried I might be turning into a total eccentric – due to over-exposure to self-help books, maybe, or just as a natural progression from being a partial eccentric – I check in with Seth Roberts, an American psychology professor based in China. After rigorous self-experimentation, he recently concluded that eating a 60g half-stick of butter every day made him 5% faster at mental arithmetic. Previously, he'd improved his sleep by eating quantities of pork belly each night. But if Roberts's single-mindedness is a little scary, it's surely also admirable: he approaches his body like a scientist, pushing aside personal biases, adjusting inputs and measuring outputs to see what influences his happiness and health.

On a more modest level, even those of us who couldn't stomach buttered butter for breakfast like to think we deal with life this way. Faced with any choice, especially big ones, we use our rational minds to identify reasons for and against, test them if possible, then do what seems most sensible. We know we're not infallible: numerous biases lead us astray, and we're horribly prone to rationalisation – that is, misusing our reasoning faculties to corral our emotions into line. But these are exceptions, we tell ourselves. After all, we're rational beings. That's what separates us from horses, or sardines, or Jeremy Clarkson.

Yet a forthcoming paper by the cognitive scientist Dan Sperber and the philosopher Hugo Mercier, Why Do Humans Reason?, proposes a radical alternative. What if we evolved the capacity to reason not to get closer to the truth, but to persuade others (and ourselves) of viewpoints, regardless of their relation to truth? In evolutionary terms, the survival benefits of such a talent are obvious. Maybe – to borrow the analogy used by Jonah Lehrer, who highlighted the paper on his blog at wired.com/wiredscience – we don't go about life as quasi-scientists, as we flatter ourselves, but as quasi-talk radio hosts, devoting our reasoning energies to concocting arguments that feel persuasive.

This is speculation, but Sperber and Mercier show it makes sense of countless psychological quirks that otherwise seem mysterious. Lehrer cites the famous study in which people were asked to rate five jams previously rated by food experts. Non-experts ranked them the same as experts – except those who were asked to provide reasons, who diverged hugely, preferring jams that (according to expert opinion) were worse. Seemingly, they were casting about for convincing-sounding reasons – "Smoother jam is better", say – which threw them from their instinctive preference for the jams everyone else agreed were best. If reasoning is about truth-finding, this is bewildering, but if it's about generating fuel for persuasion, it makes sense. Rationalisation, from this perspective, isn't a failure of reasoning. It's what reasoning's for.

If so, the implications are big. It hints that reasoning your way to a decision, or being persuaded by others' arguments, is unlikely to be better than trusting your gut. And it suggests that to engage in self-experimentation like Seth Roberts is a constant battle against biases that aren't just bugs in your mental software but, rather, fundamental to how it works. You can probably think of many persuasive reasons why eating half a stick of butter a day isn't sensible. But just because they're persuasive, does that makes them more likely to be true?

oliver.burkeman@theguardian.com
twitter.com/oliverburkeman