Preparing a decision with AI: what information stays out of the briefing?
An AI-assisted summary can give a manager a coherent starting point without incorporating observations scattered across different functions. The practical question is precise: how can information held by only one person enter the decision process before the shared briefing sets the terms of the debate?
What the literature establishes
Garold Stasser and William Titus (1985, Journal of Personality and Social Psychology) demonstrated experimentally that group discussions can favour information already known to several participants over information held by only one member. In their study, this incomplete pooling could prevent the group from identifying the option favoured by the full set of available information. Information being available somewhere within a group therefore does not guarantee that it will actually inform the decision.
The mechanism: shared information has more routes into the discussion
Garold Stasser and William Titus (1987, Journal of Personality and Social Psychology) investigated how information load and the proportion of shared information influence the dissemination of unshared information. A central mechanism is sampling: information known to several people has more opportunities to be introduced into the discussion. A shared AI-generated summary does not, in itself, remove this asymmetry.
The premature conclusion
One might conclude that AI should simply be asked to produce a more comprehensive summary. Yet a document’s completeness is not the same as the completeness of the knowledge distributed across a team: a shop-floor observation missing from the supplied data cannot be recovered through a summarisation instruction alone. The management task is therefore also to organise information gathering upstream, not merely to improve the writing. (our executive and employee training programmes)
What these findings do not establish
These experiments concern group decisions in controlled settings, not contemporary AI use in Neuchâtel’s industrial companies. They demonstrate neither that an AI-generated summary worsens this bias nor that collecting information beforehand is sufficient to correct it. They justify attention to how information circulates, without providing an estimate of the effect to expect in a workplace.
A practical check in Neuchâtel
A Neuchâtel company working in microtechnology, watchmaking or precision microelectronics could test this when deciding how to handle a non-conformity: before circulating the AI summary, ask production, quality and manufacturing engineering separately to record relevant facts that may be known only to their own function. After removing duplicates, measure the proportion of these initially unshared facts explicitly examined before the decision, recording how each was considered or why it was excluded. This indicator measures whether information enters the discussion, not the final quality of the decision. Within SHR’s « Manager à l’ère de l’intelligence artificielle » programme, it offers a concrete exercise: check what the briefing brings into the debate rather than judging only its readability. To go further: explore the Managing in the Age of Artificial Intelligence training in Neuchâtel in the canton of Neuchâtel, or browse our executive and employee training programmes in Switzerland.
In pictures: Managing in the Age of Artificial Intelligence in Neuchâtel
- managing in the age of artificial intelligence
- Neuchâtel
- research
- training neuchâtel