The problem of eliciting and aggregating information arises at several levels, including the social (wisdom-of-the-crowd) and the neural (ensemble voting). Elicitation requires incentives that will reward honest reporting of private signals.  Aggregation involves selecting the best report from an entire distribution (Galton famously proposed the ‘democratic’ median, back in 1907). The lecture will present recent work on elicitation (Bayesian truth serum) and aggregation of individual judgments, in a setting where truth is not verifiable and prior information about individual competence is lacking. Non-verifiable domains encompass historical and artistic judgments, futuristic forecasts, as well as phenomenological “first person” descriptions of mental activity. The formal results use Bayesian game theory, and exploit information contained in predictions by individuals about how others will respond. Experimental results illustrate the approach.