[ESSEC Knowledge] by Estefania Santacreu-Vasut - Associate Professor of Economics at ESSEC Business School
In recent years, the #MeToo movement has brought gender inequalities to the forefront of discussions, coinciding with growing concerns about the impact of artificial intelligence (AI) on various aspects of society. AI, especially in the field of machine learning, has the potential to significantly affect the labor market, which has long been studied for gender inequalities. Researchers have examined the gender wage gap and the role of attributes and discrimination in explaining this gap. AI's ability to process vast amounts of data has raised questions about its fairness and potential to either reduce or exacerbate gender discrimination.
Three key considerations for understanding AI's impact on gender biases are highlighted. First, it's crucial to define the correct benchmark or counterfactual when assessing AI's impact. Instead of asking whether AI algorithms are prone to gender bias, the focus should be on comparing the magnitude of bias with and without AI. Human judgment is often biased, and AI should be evaluated in comparison to these human decisions.
Second, a distinction must be made between the objectives and predictions of AI algorithms. It's essential to determine whether AI goals or the actual predictions are biased. This differentiation is vital, especially for algorithms involving human oversight. Programmers and individuals in the decision-making process may carry unconscious biases that affect the algorithm's outputs.
Lastly, when implementing policies to counteract biased objectives or predictions, the approach may differ. Legal tools may be useful when dealing with biased objectives in AI algorithms. However, stringent legal measures might encourage less transparent algorithms, leading to an information gap between regulators and users. Alternatively, addressing biases in AI predictions may rely more on education and training to combat human biases and to understand that data used by algorithms can contain biases. The ultimate success of AI in addressing gender inequalities depends on tackling the root cause: human biases.
[To read the full article please follow this link.]
