If we treat current Australian AI adoption concentration as partly an artefact of who counts as a user—for example, counting each institutional licence in metro hubs but often ignoring shared or mediated use in regional settings (library-mediated use, union- or industry-body–hosted tools, community hub access)—how would a person-centred, mediated-use–inclusive metric of per-capita adoption and use-case mix change our picture of regional vs metropolitan equity, and what contradictions would that expose in existing state and federal AI deployment targets?

anthropic-australia-usage | Updated at

Answer

A person-centred, mediated-use–inclusive metric would narrow measured metro–regional gaps in per-capita AI adoption and shift the apparent regional use-case mix toward more work-like activity, exposing contradictions between equity rhetoric and hub- and licence-centric deployment targets that implicitly discount regional, shared, and low-status settings.

Directional changes in the picture

  • Per-capita adoption

    • Regional adoption would rise more than metro once we count:
      • Library, school, and TAFE lab access where multiple people share a small number of accounts/devices.
      • Union/industry-body portals and co-op or franchise tooling used by dispersed workers.
      • Community hub programs and public PC access.
    • Adoption concentration between leading states and others would still exist, but the apparent metro–regional gulf would be smaller than licence data suggest.

  • Use-case mix (work vs coursework vs personal)

    • Regional areas:
      • Work: share of use would rise once mediated access for small businesses, farm work, local services, and union-supported tools is counted.
      • Coursework: would rise modestly where schools/TAFEs give supervised lab or classroom access without individual licences.
      • Personal: relative share would fall (similar or slightly higher absolute use, but a smaller proportion of total once mediated work/course use is visible).
    • Metro areas:
      • Smaller change, because much work and coursework use is already visible via institutional licences.

  • Equity interpretation

    • The story shifts from “regions don’t use AI” to “regions use AI more than we record, but via shared, lower-status, and less supported channels.”
    • Measured inequity looks less about absolute absence of use and more about:
      • Quality and safety of tools.
      • Depth and productivity of use.
      • Whether use is institution-supported or fragile, personal/mediated access.

Contradictions this exposes in existing deployment targets

  • Licence- and hub-centric targets vs equity rhetoric

    • Targets framed around numbers of institutional licences or pilots in major universities, departments, and hospitals will overstate metro leadership and understate regional participation.
    • Equity strategies that claim to “close regional gaps” while tracking mainly hub licences and formal deployments contradict a person-centred reality where many regional users already participate via mediated access.

  • ‘Innovation hub’ focus vs everyday regional usage

    • Policies that prioritise AI hubs and precincts as primary delivery vehicles presume that innovation diffuses outward.
    • Inclusive metrics reveal parallel, under-recognised usage in regional libraries, TAFEs, community health, and small firms that are not treated as core sites in deployment targets.

  • Status hierarchy in what counts as legitimate AI use

    • University coursework and professional knowledge work are often counted and celebrated; TAFE labs, apprenticeship block-release training, and union-hosted tools are rarely central to targets.
    • A mediated-use–inclusive lens shows that many regional and VET/apprenticeship users already engage with AI, contradicting policies that treat them mainly as future beneficiaries rather than current users.

  • Public-sector deployment goals vs actual service-equity gaps

    • State and federal AI strategies may highlight metro-based pilots in health, justice, and education as evidence of progress.
    • Once person-centred, mediated use is counted, we see that regional residents often access AI-enhanced services only indirectly (e.g., via a single TAFE campus, library, or telehealth node), and many high-need services still lack supported tools—contradicting claims that pilots meaningfully address regional service inequity.

Policy implications

  • Revise metrics to count:
    • Unique users, including those accessing via shared accounts and devices.
    • Sessions and tasks per person across work, coursework, and personal categories, regardless of licence holder.
    • Regional mediated access points (libraries, hubs, union/industry platforms) as first-class deployment sites.
  • Rebalance deployment targets toward:
    • Entitlement-style offers for regional schools, TAFEs, councils, libraries, and industry bodies.
    • Service-equity and person-level outcome measures (e.g., per-capita work and learning tasks supported) rather than licence counts in hubs.