Why measure what people don’t know they want
Ask people what motivates them at work and they will tell you something coherent: autonomy, impact, recognition, a fair salary. Those answers are explicit motives, and they are real — they predict the goals people set and the jobs they apply for.
They are also incomplete. Alongside those conscious preferences sits a second system: affective preferences that develop early, operate outside awareness, and predict a different class of outcomes — spontaneous behavior, persistence, and how energizing an activity feels rather than how sensible it seems. The two systems correlate weakly at best. A person can sincerely want a leadership role while their implicit power motive is low, and the mismatch shows up later as fatigue rather than as a change of mind.
That gap is the part of my work I find most interesting. It is not that people misreport their motives; it is that self-report can only reach one of the two systems. Which raises a practical problem: the indirect measures that reach the other one are expensive to score, which is why motive research has stayed small-sample for decades. Whether large language models can change that is the question I am working on now.