When Prediction Becomes Disposition: Governing AI-Assisted Decision Making in Animal Shelters

Predictive systems can influence far more than forecasts in animal shelters. Once predictive information begins shaping access to foster placement, veterinary treatment, behavioural support, rescue outreach, transfer opportunities, promotion, additional time, or euthanasia consideration, prediction becomes part of the process producing the outcome it purports to predict. This paper uses prediction-to-disposition as a shelter-specific governance concept for examining that transition. Drawing on research concerning animal shelter outcomes, performative prediction, selective labels, automation bias, animal welfare, and workforce pressures, it proposes a consequence-sensitive approach to AI-assisted shelter decision making. The paper examines how predictive information can affect resource allocation, professional authority, animal opportunity and welfare, organisational outcomes, and the moral burden carried by shelter personnel.