IA on the job: comparing proposals without shifting the criteria?
Comparing suppliers, equipment or training services is one possible use of AI at work. The management question is specific: should criteria be set before asking an assistant to rank proposals?
What the literature establishes
Christopher K. Hsee (1996, Organizational Behavior and Human Decision Processes) shows that preferences between options can reverse depending on whether they are evaluated separately or jointly. An attribute that is difficult to assess in isolation can carry more weight when direct comparison becomes possible. Applied to AI, this finding suggests treating the format of a comparison table as part of the decision rather than merely its presentation.
When criteria follow preferences
Eric Luis Uhlmann and Geoffrey L. Cohen (2005, Psychological Science) show, in employee-selection experiments, that participants can redefine merit to favour a candidate on the basis of gender. In their experimental setting, committing to criteria before learning the candidates’ gender eliminates the discrimination observed. The relevant mechanism is that a seemingly defensible criterion can be chosen retrospectively to justify a preference.
The premature conclusion
It would be tempting to conclude that setting a framework in advance makes every comparison objective. These studies support neither that generalisation nor the claim that AI necessarily reproduces the same mechanisms. A proposal may reveal an overlooked requirement; the useful discipline is then to document the change and reassess all proposals against the revised framework. (our executive and employee training programmes)
What the final table cannot reveal
These studies concern neither generative assistants nor procurement by businesses in Fribourg. They illuminate judgement mechanisms without measuring the effect of an AI-generated ranking in this context. A final table, even with explanations, cannot on its own distinguish a criterion established beforehand from one introduced after the proposals were reviewed.
A practical check in Fribourg
As part of SHR’s “IA on the job” programme, a food-processing company, a bilingual industrial site or a team within Fribourg’s university community could test the same rule during its next comparison of non-sensitive proposals. Before giving the proposals to the assistant, the team dates and saves its criteria, any weightings, and their French and German wording approved by the relevant operational specialists. It then measures the proportion of criteria added, removed or modified after reviewing the proposals, keeping the justification for each change. Every change triggers a fresh assessment of all proposals: this record checks the stability of the comparison framework without claiming to prove that the chosen supplier is the best. To go further: explore the AI on the Job training in Fribourg in the canton of Fribourg, or browse our executive and employee training programmes in Switzerland.
In pictures: AI on the Job in Fribourg
- ai on the job
- Fribourg
- research
- training fribourg