What is Automated Decision-Making (ADM)?

Automated decision-making is the process of making decisions using automation without direct human involvement, and may involve a degree of human oversight and intervention in some uses. ADM can be found across a large and growing number of sectors, with ramifications across all walks of life, impacting on everyday lives – including public administration, welfare, health, education, law, business, employment, transportation, media and entertainment. PPA research has drawn attention to the fundamental effects of the underlying ‘logics’ that have been instrumental to the way that Automated Decision-Making (ADM) systems and models are built.

Decisions can be based on data, on digital profiles or inferred data, and require a codified set of rules and criteria to reach a decision. Fine-tuning the outcomes improves efficiency and accuracy, and creates a feedback loop that measures the systems performance against the rules and criteria, creating what is sometimes referred to as a self-learning and self-correcting system. Anomalies and exceptions are singled out for review by humans.

PPA research has drawn attention to the fundamental effects of the underlying ‘logics’ that have been instrumental to the way that Automated Decision-Making (ADM) systems and models are built. PPA researches how the decisional logic encoded into constructed ADM artefacts. Giving itself a wider remit than the ‘bias’ or ‘explainability’ or artefacts, PPA has considered the effects of the social processes that occur during the design, implementation (and to a lesser degree) operation of ADM systems.

PPA has honed in to domains such as refugee resettlement, health and social care, and has worked to surface the logics associated with ADM. This provides people –stakeholders and our other participants – with a means of engaging with decisions concerning these background logics, and of identifying any unexpected outcomes that might emerge from them.