LAUSR

Affect detection from wearables in the “real” wild—where people go about their daily routines in heterogeneous contexts—is a different problem than affect detection in the lab or in the “quasi” wild (e.g., curated or restricted contexts). The U.S. government recently supported a program to develop and evaluate the performance of contemporary affect detection systems in the real-wild along the dimensions of accuracy, robustness, and generalizability. Evaluations by an independent testing team revealed that none of the performing teams met the aspirational performance metrics. Alarmingly, performance was near zero for several cases. This article is the result of soul searching to reconcile the chasm between expected and achieved performance in light of past successes of the field. We discuss the major challenges faced, their implications for future research, and suggest a path forward.