Bayesian ARgumentation via Delphi

  • Belton, Ian (Co-investigator)
  • Bolger, Fergus (Principal Investigator)
  • Crawford, Megan Michelle (Co-investigator)
  • Hamlin, Iain (Co-investigator)
  • MacDonald, Alice (Co-investigator)
  • Rowe, Gene (Principal Investigator)
  • Sissons, Aileen (Co-investigator)
  • Taylor Browne Lūka, Courtney (Co-investigator)
  • Vasilichi, Alexandrina (Co-investigator)
  • Wright, George (Principal Investigator)

Project: Research

Project Details

Description

BARD was a 23-month project funded by the US Government Intelligence Advanced Research Projects Activity (IARPA) and formed part of the larger Crowdsourcing Evidence, Argumentation, Thinking and Evaluation – “CREATE” – program. In BARD, we designed and produced Graphical User Interfaces (GUIs) to assist in the construction of Causal Bayesian Networks (CBNs) as the underlying engines for the analysis of arguments and evidence. BARD thus allows analysts to build and test competing or complementary arguments, and to examine the impact of different pieces of evidence, in an intuitive environment. BARD makes use of the Delphi technique – an iterative survey method that minimizes negative effects of cognitive and social biases – to manage the interaction between users.

In addition to the Delphi Team based in Strathclyde, BARD also consisted of teams based in London (UCL and Birkbeck) – who are experts on the psychology of causal reasoning – and in Melbourne, Australia (Monash University) – who are expert in CBNs and software engineering.

IARPA - https://www.iarpa.gov/
CREATE - https://www.iarpa.gov/index.php/research-programs/create

Layman's description

We designed and tested an online system that helps improve group decision-making when presented with uncertain evidence.

Key findings

- We developed an online system that permits collaborative construction of CBNs with as little as 2 hours of training.
- We discovered several effective methods that help both groups and individual participants break down complex problems and identify key components necessary for reaching better solutions.
- We also developed and formalised a novel approach to studying group reasoning processes, the Simulated Group Response Paradigm (SGRP).
Short titleBayesian ARgumentation via Delphi
AcronymBARD
StatusFinished
Effective start/end date1/04/1730/11/18

Keywords

  • Bayes nets
  • Delphi
  • GUI
  • Natural Language Generation
  • Cognition
  • Forecasting
  • Group judgment