Facial recognition and responsibility in public art
Public art often occupies sites where people expect to walk, wait, celebrate or pass through without becoming data. That makes the ethics of using facial recognition in public art more than a software question: it is a question of power, attention and who may remain unobserved. ITP 30 Show’s history of interactive experiments gives artists a useful vantage point, but experimental intent does not remove obligations to people caught in an installation’s field of view.
Not every camera-based work performs biometric identification. Face detection, facial analysis, identity verification and recognition have different capacities and risks. Ethical practice begins by naming the exact function, documenting what is inferred and deciding whether identification is necessary at all. Scale and setting can change the answer: a playful workshop and a citywide installation do not create equivalent exposure.
Biometric art is never just technical
A camera that identifies gallery visitors can turn a shared encounter into a ranking, a record or a means of exclusion. It can also expose the source and limits of a training dataset, including whose faces were overrepresented. The ITP 30 Show’s featured creators demonstrate how interactive media can interrogate systems rather than simply deploy them; their credits also offer a model for naming the people responsible for research, development and display.
Artistic provocation does not automatically offset harm. A work labelled satire may still make invasive identification seem normal, while a participatory performance may conceal a power imbalance beneath an invitation to “play”. Artists should separate a simulated recognition system from one that makes real-world identity claims, and state clearly whether data can identify a person, infer a protected trait or only estimate an expression.
Australian settings change the ethical questions
Australian cities combine ordinary digital habits with expanding security infrastructure. Sydney commuters tap Opal cards at transport gates, Melbourne passengers use Myki on trains and trams, and people moving through Brisbane CBD or festival precincts pass layered CCTV. Banks, retailers and government contractors also sell or use identity checks. These familiar systems matter because a public artwork may enter the same civic ecology of convenience, suspicion and assumed acceptance.
There is no single federal facial-recognition statute. Under the Privacy Act 1988, biometric information used for automated identification or verification is generally treated as sensitive information when the Privacy Act applies, while government agencies operate under other regimes. State and territory surveillance-device laws add different rules for cameras and monitoring. Australian Privacy Commissioner guidance, privacy impact assessments and the Australian AI Ethics Principles give artists practical reference points, but none replaces clear governance or a community-centred decision.
Consent must extend beyond the participant
Consent in a public place is easily mistaken for permission. A small sign at the entrance cannot explain a complex model, and a ticket or volunteer form rarely makes refusal costless when the artwork is tied to access, employment or belonging. Children, people with limited English and those accustomed to institutional surveillance may feel less able to decline. Meaningful consent requires comprehension, alternatives and a route that does not punish refusal.
That principle extends to people who never intended to participate. Notices, spacing, gaze direction and audio cues can shape who feels observed. The show’s interactive sound design journal piece offers a useful contrast: sensory feedback can make an installation responsive without identifying a body. Where recognition is retained, projects should limit collection, prohibit reuse and set deletion dates. Aboriginal and Torres Strait Islander communities also require direct engagement on surveillance, cultural safety and data sovereignty, not generic consultation.
Recommendations for accountable installations
An accountable project treats governance as part of the artwork’s interface. Its documentation should identify the system’s operator, purpose, legal basis, data flows, error conditions and accountable decision-makers in language visitors can understand. Curators should ask what the project would look like if biometric identification were removed, then test whether that alternative still supports the artistic idea.
- Publish a plain-language notice at every entrance and online, naming the system’s purpose, operator, data categories and contact route.
- Make participation voluntary and establish an equivalent non-biometric route that does not separate or disadvantage anyone.
- Complete a privacy and equity impact assessment, testing error rates and likely effects on distinct Australian communities rather than relying on aggregate accuracy.
- Set short retention periods, prohibit secondary use and document deletion, including backups and vendor copies.
- Commission independent legal, security and community review, then publish the resulting changes.
These measures should appear in commissioning agreements, artist contracts and vendor terms, not only in a curatorial essay. They also need a visible complaint process and an incident plan for system failure, data leakage or discriminatory outcomes. Review should continue after opening, when real audiences reveal assumptions that prototypes could not.
Keep scrutiny open after opening night
Accountability continues after opening night. Audiences should be able to read the data policy, contact a named steward and learn what was deleted, corrected or refused. For enquiries involving the show, its contact channel provides a direct route to the ITP 30 Show team. Project pages can similarly foreground consent terms and the limits of each system.
Organisers should publish those limits before promotion, invite scrutiny during development and keep a clear correction channel open throughout the event. Artists should credit technical collaborators, fund community review and avoid claiming that creativity transfers responsibility away from them. Read the project notes, demand a visible no-surveillance alternative and support only installations whose data promises can be tested in public.