Beyond Distributive Justice: Hermeneutical Fairness in Ad Delivery
Ad algorithms may cause invisible harm by distorting how people understand the world.
Fairness in online advertising has traditionally focused on distributing ad impressions or outcomes equally across demographic groups—a concept known as distributive justice. Yet, as researchers from Politecnico di Milano argue in a new preprint (arXiv:2605.03419), this approach misses a critical dimension: how ad content shapes recipients' understanding of the world. Drawing on Miranda Fricker's notion of hermeneutical injustice, they propose that ads act as 'interpretative resources' that can be withheld (causing hermeneutical deprivation) or skewed through saturation with misleading frames (producing hermeneutical distortions). Using exploratory patterns from AIDS Advertising Evaluation surveys (1986-1987) as a case study, they demonstrate that purely economic optimization systematically under-delivers critical public health messages to marginalized groups, even when distributional parity appears fair.
To address this, the team introduces a group-level hermeneutical fairness constraint and a hermeneutically aware utility cost, integrated into a benchmark ad allocation framework that already enforces distributive justice. Controlled simulations reveal a nuanced trade-off: purely utility-based allocation drives under-delivery to the disadvantaged group, while incorporating distributive constraints significantly reduces hermeneutical cost at modest utility loss. However, weighting hermeneutical cost alone (without distributive constraints) can lead to policies that concentrate ads heavily on one group, creating new inefficiencies. The findings advocate for expanding fairness analyses beyond just 'who sees what' to include the epistemic conditions of interpretation and uptake—a timely warning for ad platforms and policymakers as AI-driven ad systems become more pervasive.
- Proposes hermeneutical fairness as a new dimension beyond equal distribution of ad impressions
- Uses 1986-87 AIDS advertising data to show utility-driven systems under-deliver to disadvantaged groups
- Combined distributive + hermeneutical constraints reduce epistemic harm with only modest utility loss
Why It Matters
Ad platforms must now consider not just who sees ads but how content shapes understanding—critical for public health and politics.