AI adoption: how can work be assessed fairly?
When some employees can delegate part of their work to AI and others cannot, the same productivity expectation may conceal very different working conditions. The precise question for a team manager is this: how can work be assessed without turning unequal access to AI into a judgement about individual merit?
What the literature establishes
Jason Colquitt (2001, Journal of Applied Psychology) provides empirical support for four distinct dimensions of organisational justice: distributive, procedural, interpersonal and informational. Fairness therefore concerns not only the outcome of an assessment, but also its procedure, the treatment of people and the explanations provided. Applied to AI, this distinction calls for separate consideration of what is rewarded, how the decision is made and how it is justified.
Explanation is more than a supporting message
Jerald Greenberg (1990, Journal of Applied Psychology) studied reactions to a temporary pay reduction in an industrial setting. His field study indicates that a thorough explanation attentive to employees’ circumstances can mitigate adverse reactions, measured in particular through theft. It does not concern AI, but suggests that explaining a new assessment rule should not be treated merely as a communication exercise.
The hasty conclusion
One might conclude that clearly explaining an AI-related increase in targets is enough to make it fair. That would confuse the quality of the explanation with the quality of the rule: a transparent criterion can still be inappropriate when tasks, access rights or confidentiality obligations differ. The question is therefore not simply how to secure acceptance of an assessment, but whether the contributions being compared were made under sufficiently comparable conditions. (our executive and employee training programmes)
What these findings cannot determine
These studies establish neither a method for assessing AI-assisted work nor a causal effect of fairness on the adoption of such tools. Nor do they determine how much of a result should be attributed to the employee, the software or the resources provided. In practice, raising targets on the basis of a few quickly completed outputs leaves verification work and restrictions on use out of the picture.
A practical check in Sion
As part of SHR’s « Conduire l’adoption de l’IA dans son équipe » programme, a team in Sion could spend one month examining every performance target changed on the grounds of AI use. For each target, it would record the tasks concerned, authorised access to the tool, expected verification time and justified exceptions: confidentiality in healthcare, technical validation in hydropower, or seasonality in viticulture and Alpine tourism. The measure would be the proportion of these targets for which those conditions are documented and communicated before assessment, retaining both the numerator and the denominator. This record would not prove fairness, but would make one prerequisite verifiable: people should not be judged against expectations whose conditions of application remain implicit. To go further: explore the Leading AI Adoption in Your Team training in Sion in the canton of Valais, or browse our executive and employee training programmes in Switzerland.
In pictures: Leading AI Adoption in Your Team in Sion
- leading ai adoption in your team
- Sion
- research
- training sion