The CACHE study: group effects in computer-supported collaborative analysis

Abstract

The present study evaluates some effects of group composition in distributed intelligence analysis. Prior work suggests that bias makes this task difficult. We experimentally investigated the impact of heterogeneity of view point and prior knowledge on bias and information coverage, specifically on anchoring and confirmation bias. Bias was substantially reduced whether individuals worked alone or in heterogeneous groups yet, importantly, net process costs were not increased for collaborating individuals. Also, despite similar initial bias, individuals in homogeneous groups showed increasing bias and greater final bias than in the other conditions suggesting that ill-formed groups exacerbate bias. All participants used CACHE, a web-based environment that supports a collaborative version of Analysis of Competing Hypotheses (ACH)a method employed by professional intelligence analysts. We suggest that CACHE may be effective in reducing both bias and process costs in collaborative analysis.