Sociotechnical harms of algorithmic systems: Scoping a taxonomy for harm reduction
A structured algorithmic harm taxonomy gives APS governance practitioners a reusable vocabulary for identifying and categorising AI-related risks in agency contexts.
Key points
- A 2023 AAAI/ACM paper taxonomises algorithmic harms into five categories and 20 subcategories.
- Categories span representational, allocative, quality-of-service, interpersonal, and social system harms across micro to macro levels.
- Spotlighted via MIT AI Risk Repository as a reference framework - useful for harm identification but not new or AU-specific.
Summary
This MIT AI Risk Repository blog post spotlights a 2023 peer-reviewed paper that presents a taxonomy of sociotechnical harms from algorithmic systems, derived from a scoping review of 172 computing research papers. The taxonomy organises harms into five major categories - representational, allocative, quality-of-service, interpersonal, and social system harms - each with subcategories. It frames impacts at micro, meso, and macro societal levels. The taxonomy is an applied reference tool for researchers and practitioners seeking shared language to identify and reduce AI-related harms.
Implications for Australian agencies
- Consider APS teams developing AI risk registers or harm assessment frameworks could consider adopting or mapping against this taxonomy to ensure coverage across harm types.
- Monitor Agencies tracking the MIT AI Risk Repository may want to monitor additional frameworks spotlighted in this series for complementary risk perspectives.
Implications are AI-generated. Starting points, not advice.
"Sociotechnical harms of algorithmic systems: Scoping a taxonomy for harm reduction" Source: MIT AI Risk Repository – Blog Published: 16 January 2025 URL: https://airisk.mit.edu/blog/sociotechnical-harms-of-algorithmic-systems-scoping-a-taxonomy-for-harm-reduction This MIT AI Risk Repository blog post spotlights a 2023 peer-reviewed paper that presents a taxonomy of sociotechnical harms from algorithmic systems, derived from a scoping review of 172 computing research papers. The taxonomy organises harms into five major categories - representational, allocative, quality-of-service, interpersonal, and social system harms - each with subcategories. It frames impacts at micro, meso, and macro societal levels. The taxonomy is an applied reference tool for researchers and practitioners seeking shared language to identify and reduce AI-related harms. Implications for Australian agencies: - [Consider] APS teams developing AI risk registers or harm assessment frameworks could consider adopting or mapping against this taxonomy to ensure coverage across harm types. - [Monitor] Agencies tracking the MIT AI Risk Repository may want to monitor additional frameworks spotlighted in this series for complementary risk perspectives. Retrieved from SIMS, 18 May 2026.