Companies are increasingly using AI to calculate salaries, bonuses, and other employee payments. Algorithms can consider more factors, compare compensation with the market, and identify imbalances that a human might miss, said Alexander Khaminsky, head of the Center for Law and Order in Moscow and Moscow Oblast.
However, the objectivity of such a system depends on the data and rules embedded in it. If the initial criteria inherently contain biases, AI may not correct them but merely spread them to a larger number of employees.
An algorithm does not determine fairness independently; it merely reproduces the embedded criteria. If these are incomplete or reflect existing discrimination, AI may not eliminate it but, on the contrary, automate and scale inequality.
At the same time, the final personnel decision cannot simply be delegated to an algorithm. The employer remains responsible for complying with labor laws, ensuring equal pay for work of equal value, and preventing discrimination.
An employee must also understand why their salary was changed or why they did not receive a bonus. The phrase "AI decided so" cannot by itself be considered a sufficient explanation. Therefore, a neural network here is more of an analysis tool than an independent arbiter.