IA on the job: can one extra option change the choice?
Generative assistants can quickly prepare a shortlist of providers, training courses or technical solutions. The management question is specific: how can we check that a choice reflects the merits of the offers rather than the presence of an option that makes another look better?
What the literature establishes
Joel Huber, John W. Payne and Christopher Puto (1982, Journal of Consumer Research) showed that adding an alternative dominated by one option, but not by another, can increase the probability of choosing the option that dominates it. An apparently inferior offer can therefore change the trade-off between the original offers. Choice depends not only on each proposal’s characteristics, but also on the set presented.
The mechanism: making a preference easier to justify
Itamar Simonson (1989, Journal of Consumer Research) examined attraction and compromise effects in relation to the reasons available to justify a choice. An option can become more attractive because it easily outperforms another, or because it occupies an intermediate position. Applied to an AI-assisted comparison, this mechanism suggests that the composition of the shortlist can provide a convenient justification without supplying new information about suitability.
The premature conclusion
One might conclude that weaker offers should always be removed, or that only two possibilities should be retained. These studies do not warrant such a rule: an offer that performs less well on some criteria may meet a constraint omitted from the table. The useful precaution is to distinguish criteria established before the comparison from arguments that emerge because a particular option appears on the list. (our executive and employee training programmes)
What the evidence cannot establish
These studies concern experimental choice situations, not the deployment of generative assistants in Geneva businesses. They establish neither the frequency nor the magnitude of these effects in professional procurement, where contractual constraints and multiple decision-makers play a role. They support checking sensitivity to the comparison context, not claiming that every AI-generated ranking is manipulative.
A practical check in Genève
Within SHR’s “IA on the job” programme, one possible exercise uses a fictional provider-selection case adapted to private banking, an international organisation, luxury watchmaking or commodity trading. Randomly assign participants to two versions of the same comparison: the main offers and their characteristics remain identical, but only one version includes an offer dominated on the criteria established beforehand. Measure the percentage-point difference between the proportions choosing the offer that dominates this alternative in the two versions. This difference is a local signal to investigate, not general proof; above all, the exercise trains managers to examine the composition of a comparison before defending its conclusion. To go further: explore the AI on the Job training in Geneva in the canton of Geneva, or browse our executive and employee training programmes in Switzerland.
In pictures: AI on the Job in Geneva
- ai on the job
- Geneva
- research
- training geneva