Carissa V9liz Frames AI Predictions as Ethical Problem

EL PAdS reports that Oxford University philosopher Carissa V9liz has published a book titled Prophecy that examines how statistical and probabilistic reasoning  now widely embedded in artificial intelligence systems  functions as a mechanism of power. In an EL PAdS interview, V9liz is quoted saying, "Predictions are often commands disguised as a search for knowledge," and she argues that a probabilistic mindset is "emerging alongside AI and presents predictions as facts," which she describes as having "profound ethical implications." The article also notes V9liz previously wrote Privacy Is Power and advises Spain's Ministry for Digital Transformation and Public Administration on AI matters, per EL PAdS.

Industry-pattern observations: Modern machine learning systems produce calibrated probabilities, scores, and ranked outputs rather than categorical truths. When those probabilistic outputs are communicated to nontechnical users without uncertainty framing, they are often perceived as definitive answers. This mismatch between model epistemic status and public reception is a recurrent issue in model deployment, affecting trust, decision-making, and downstream behavior across domains from hiring to policing.

Editorial analysis: The framing V9liz raises intersects with ongoing policy debates about transparency, explainability, and informed consent. Policymakers and regulators increasingly confront cases where automated predictions materially alter people's lives; the ethical concern V9liz highlights is that probabilistic outputs can shape expectations and, through those expectations, influence outcomes. That dynamic amplifies questions about responsibility, contestability, and the social consequences of automation.

For practitioners: monitor how product teams, compliance functions, and regulators translate probabilistic model outputs into user-facing language, documentation, and appeals processes. Observers should track efforts to standardize uncertainty communication, audit predictive pipelines for feedback loops, and require impact assessments where predictions carry legal or social weight.