Sensemaking in a largescale corporate fraud investigation The
- Slides: 8
Sensemaking in a large-scale corporate fraud investigation The communication challenge (and four problems of search) Simon Attfield Ann Blandford John Dowell Making Sense of Information Project EPSRC Grant EP/D 056268 (June 2006 - May 2009)
Context • Making Sense of Information project (lawyers and journalists) • Sensemaking is… Ø Sensemaking is a process through which people structure information to form an integrated view about something. Ø Data-frame (model) symbiosis (Klein et al, IEEE Intell. Sys. 2006) Ø A frame supports rapid convergence on new information Ø Sensemaking is an antecedent of action (recognition primed decision making)
Legal Investigation Case-study • A large corporate fraud investigation • In depth, retrospective interviews with 9 out of 30 investigating lawyers (45 min to 1 hr 40) • Aim for rich accounts of: – Information interactions (filtering, reviewing, integrating, sharing) – How integrated representations supported reasoning • This talk… – Brief overview – Collaboration – Some search challenges
The case • Regulatory context • Objective: assess specific allegations and related vulnerabilities • Data-forensics recovered 8. 5 million paperbacks-worth (30% emails) Search
The integrated representation The event corroboration problem The structuring the information space problem
The Communication Challenge “The flow of information around the group, is one of the most challenging parts of it. ” • Problem setting (constant review) – Explicitly defined issues and questions – Propagating coding knowledge • Sharing findings/expressing interests – “You have to move information around” (duplicates problem) – “You’re constantly having to be aware of other peoples interests” (Interest recognition problem) • Challenging – Document linking
Solutions? • Event corroboration – Date reference normalisation and search? – Visual representation? • Structuring the information space – Ranking by generality (number of date references, term coverage? ) • Duplicates (indication to not share? ) – (Near) duplicate match between new documents and reviewed documents? • Interest recognition (indication to share) – Similarity match between new documents and linked documents?
End
