Until a few weeks ago, I relied on restaurant hygiene ratings when deciding where to have dinner. The score 鈥� given by the UK鈥檚 Food Standards Agency 鈥� ranges from 0 to 5, where 0 indicates urgent improvement is required. It is traditionally a good predictor if you are fastidious 鈥� or if you just want to avoid nasty parasites like .
Yet in October, the BBC some businesses misrepresent their hygiene scores, posting a 5 when their ratings are as low as 0. The signal of hygiene I relied on wasn鈥檛 always reliable.
Advertisement
This was just one incident that had an impact on my trust. Yet it represents something older and more pernicious: deception through misrepresentation undermines trust. As with any social behaviour, when betrayals are successful, they tend to spread, because people behaviours associated with success. Over time, this may erode the shared signals on which society depends.
Research in biology and sociology highlights how. Across the natural world, animals signal information to other organisms about themselves: dogs to display a potential for aggression, for example, and male red deer to indicate their size.
Some animals, however, cheat. Cuckoo chicks, for instance, the chirps of the young of other birds to deceive adult birds into feeding them. This dishonest behaviour evolved because it works 鈥� in an analogous way to how deception spreads, whether biologically or, in the case of humans, culturally.
Among humans, technological improvements allow us to misrepresent our intentions and identities in many ways. This can include masking who we are in digital interactions by 鈥渃atfishing鈥� and stealing from people online without ever meeting them.
Generative artificial intelligence is only increasing opportunities of this kind. In the book , sociologists describe ways AI can be used to deceive people in hostile interactions, even in wars. Early use of voice manipulation, for example, permits people to pose as senior military officials 鈥� or even your . We share a lot of qualities with cuckoos: anyone can be anyone in the digital age.
Luckily for us, we have found ways to solve problems with trust throughout our history. This is because humans are good at finding out who is trustworthy over time. We learn to recognise patterns of behaviour indicative of deception, and just as the tactics for deception spread, so do the methods for , whether those involve passwords, face or voice recognition, or any of the new technologies being developed to combat deepfakes.
People are also adept at developing norms that place a cost on deceivers. We have created laws that impose fines or imprison those guilty of impersonation. We are only at the early stages of developing effective rules around AI, but there is a reason we are focusing on the dangers these tools pose to society. That awareness is itself a step in our defence.
What matters for retaining trust is the reliability of the signals we use in communication. It doesn鈥檛 matter whether we are talking about hygiene scores or phishing emails. What we need is a mechanism for determining reliability: whether people or places can be trusted to do what we expect of them. This is done both through effective policing 鈥� sanctioning those who show themselves to be untrustworthy 鈥� and through educating ourselves about how to avoid , places and things. If I can trust that hygiene ratings are reliable, then I can use the score as a signal to drive my behaviour.
We need to adapt how we learn to place trust and educate people how to spot deception. Even those of us concerned with the accuracy of food hygiene ratings.
Jonathan R. Goodman is the author of Invisible Rivals: How we evolved to compete in a cooperative world (2025)
