Hypothesis generation in investigative contexts

• Main Author: Sandham, Alexandra Louise
• Published: Lancaster University 2012
• Subjects: 153.8
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.656863

The generation of decisions concerning responses to uncertain evidence is a significant source of errors in investigations. In an attempt to understand the basis of investigative decision-making and illuminate the basis of flawed decisions, this thesis investigated how individuals generate hypotheses to explain uncertain evidence. The hypotheses generated by novices and experienced investigators were explored, and subsequently those generated by dyads and triads. The theoretical framework of mental models theory was used to explore the effects of event type and event familiarity on the number and types of hypotheses generated for investigative scenarios where the evidence presented is open to question. In an attempt to improve hypothesis generation performance, training based upon analogical transfer and the use of visual support aids were also evaluated. Analysis of numbers and types of hypotheses generated showed that both laypeople and experienced investigators struggled to generate those hypothesis types identified by mental models theory as difficult to discover. In addition, all participants showed an evidential truth bias with familiar events. Whilst training based upon analogical reasoning increased the number of all kinds of hypothesis generated and removed biases associated with familiarity, training did not help all participant types to consistently generate all hypotheses types across all event types. These results show that multiple hypothesis generation, although crucial in investigative contexts, is difficult and intransigent. Contrasting effects of collaboration were found. Triads generated more hypotheses than individuals who in turn reported more hypotheses than dyads. Results show support for theories of inhibition based upon collaborative satisficing for dyads and facilitation for triads based upon falsification by a third party. The results are interpreted within a putative process model of investigative hypothesis generation. Initial hypotheses are based on the construction of a mental representation of evidence and implications guided by the operation of a principle of truth, with subsequent fleshing out to discover alternative hypotheses. Hypothesis generation is further moderated by inclusion of hypotheses made available from familiarity with scenarios, and by exclusion of hypotheses depending upon the goals of the reasoner. Collaborative performance further impacts on hypothesis generation ·1 by imposing either strategies of satisficing that limit search for alternative hypotheses or of falsification that enable the discovery of alternative hypotheses.