Alignment Debt: The Hidden Work of Making AI Usable
Cumi Oyemike, Elizabeth Akpan, Pierre Hervé-Berdys
TL;DR
Frontier LLMs often assume high-resource conditions that do not hold in Global South contexts, creating alignment debt—a measurable user-side burden from cultural, linguistic, infrastructural, epistemic, or interaction misfit. The authors propose a four-type taxonomy (Cultural/Linguistic, Infrastructural, Epistemic, Interaction) and validate it through a cross-sectional survey (N=411; measurable N=385) in Kenya and Nigeria, revealing prevalences of $51.9 ext{ extpercent}$, $43.1 ext{ extpercent}$, $33.8 ext{ extpercent}$, and $14.0 ext{ extpercent}$ respectively, with a mean cumulative burden of $1.43$ types per user. Epistemic debt strongly correlates with verification propensity ($91.5 ext{ extpercent}$ vs $80.8 ext{ extpercent}$; $p = 0.037$) and cumulative debt increases verification intensity ($ ho = 0.147$, $p = 0.004$), while infrastructural and interaction debts show weaker or no clear links to verification, indicating some burdens cannot be mitigated by simply verifying outputs. The paper contributes an empirical, cross-cultural instrument and taxonomy for measuring user burden, offers design and governance implications to reduce burden (e.g., low-bandwidth modes, localized sources, and uncertainty cues), and argues for context-aware safeguards in African AI governance. Overall, it reframes fairness as “who pays” for usability, urging stakeholders to invest in local data infrastructures, context-sensitive interfaces, and policy mechanisms that redistribute alignment work away from marginal users.
Abstract
Frontier LLMs are optimised around high-resource assumptions about language, knowledge, devices, and connectivity. Whilst widely accessible, they often misfit conditions in the Global South. As a result, users must often perform additional work to make these systems usable. We term this alignment debt: the user-side burden that arises when AI systems fail to align with cultural, linguistic, infrastructural, or epistemic contexts. We develop and validate a four-part taxonomy of alignment debt through a survey of 411 AI users in Kenya and Nigeria. Among respondents measurable on this taxonomy (n = 385), prevalence is: Cultural and Linguistic (51.9%), Infrastructural (43.1%), Epistemic (33.8%), and Interaction (14.0%). Country comparisons show a divergence in Infrastructural and Interaction debt, challenging one-size-fits-Africa assumptions. Alignment debt is associated with compensatory labour, but responses vary by debt type: users facing Epistemic challenges verify outputs at significantly higher rates (91.5% vs. 80.8%; p = 0.037), and verification intensity correlates with cumulative debt burden (Spearmans rho = 0.147, p = 0.004). In contrast, Infrastructural and Interaction debts show weak or null associations with verification, indicating that some forms of misalignment cannot be resolved through verification alone. These findings show that fairness must be judged not only by model metrics but also by the burden imposed on users at the margins, compelling context-aware safeguards that alleviate alignment debt in Global South settings. The alignment debt framework provides an empirically grounded way to measure user burden, informing both design practice and emerging African AI governance efforts.