Annotator Positionality as Signal: Psychometric Weighting for Anti-Autistic Ableism Detection
arXiv:2605.26397v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used in decision-making tasks where they can amplify or suppress perspectives, raising concerns in high-stakes settings affecting autistic communities. While previous research has identified disability-related biases in LLMs, it remains unclear how they conceptualize ableism or detect it in text. We introduce a bias-aware evaluation framework targeting anti-autistic ableist language with a psychometrically-weighted, community-proximate ground truth anchored in annotator positionality. This framework c
As large language models become ubiquitous in decision-making, the urgency to address their inherent biases, particularly concerning vulnerable communities, has intensified.
This work introduces a much-needed robust framework to detect and mitigate anti-autistic ableism in LLMs, directly impacting ethical AI development and societal equity.
The ability to systematically identify and measure ableist biases in AI using psychometrically-weighted, community-proximate data shifts from ad-hoc detection to a structured, evaluative approach.
- · AI ethicists and developers
- · Autistic communities
- · Responsible AI companies
- · Human rights organizations
- · Developers ignoring bias
- · Models with unmitigated biases
- · Platforms amplifying discrimination
This framework will enable more ethical and inclusive AI systems that are less prone to perpetuating harmful stereotypes.
Increased demand for specialized datasets and annotation methodologies will emerge, fostering new companies and research areas focused on AI fairness and bias mitigation.
The success of this approach for ableism could set a precedent for detecting and mitigating other forms of discrimination, leading to a broader overhaul of AI bias detection across various social dimensions.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL