Transcription
All data is unrepresentative. You know, it'll either have something like quantification bias; it'll be weighted towards what's numerical. Um, it's worth noting that what humans think and say isn't representative of what they feel. Okay?
So, an AI may completely understand what you might call what economists call stated preference, which is what people say they want. But revealed preference, which is what people actually do, which is much more the product of kind of tacit emotions and um, actually literally, I mean, uh parts of the brain that are actually opaque to introspection. So, you know, we feel these things, we don't even think them, and therefore we never say them, so they never appear in paper.
And of course, you're in a field, interestingly, in terms of staff incentives, which is actually, when you think about it, very, very psychologically determined, isn't it? Because if you go to staff and say, "What sort of staff incentive would you like? What would you find motivating?" they'll almost all say, "More money." Okay? And yet, if you actually look at what works in terms of staff retention, what fosters culture, what makes people feel valued, it's actually almost everything except money.