Noveck, B.S., Biddle, N., and Fischer, A. (2026). Learning to Listen: Why Public Engagement Is the AI Skill Australia’s Governments Are Missing. Australian Resilient Democracy Network Discussion Paper 20, Australian National University.
Full paper available here
Two trends are unfolding in parallel around the world. Public anxiety about artificial intelligence is growing while communities’ satisfaction in how they participate in civic activities and contribute to government policy and programs is declining. Pathways to resolve this tension might be where we least expect them: governments finding ways to use these technologies to listen better and enable communities to participate more meaningfully. Examples from Bogotá, Hamburg, Brazil, Mexico and elsewhere show how AI can improve the quality and experience of public engagement while reducing its time and cost. If done poorly, or without safeguards, it can also introduce new risks. However, these can be managed with appropriate process design, regulation and delivery by skilled practitioners.
This paper examines how artificial intelligence and digital platforms are changing what is possible and suggests that a key opportunity is to build public-sector capability to deliver meaningful engagement.
Key findings
Several findings emerge after reviewing the latest surveys, global case studies and policy context.
First, meaningful public engagement does not happen by accident.
This is a professional capability that most public servants have never been systematically taught, despite the importance of collective intelligence for improving decisions and solving problems. Wanting to do an engagement is not the same as knowing how.
Second, AI and digital platforms make large-scale public engagement feasible in new ways.
AI and digital platforms make large-scale public engagement feasible in new ways that were previously impractical. In the past, without ways to capture and synthesise comments from large number of participants, institutions have been overwhelmed. In those situations, the more people participate, the harder it becomes to extract signals from noise. Digital tools are now being designed and used responsibly to help engage at large scales and faster paces, supporting democratic values, rather than as a substitute for them.
Third, the current constraint is no longer technology but capability.
Governments need public professionals with the skills to design, run, and learn from meaningful public engagement, supported by the infrastructure and leadership to put those skills into practice.
The full Discussion Paper [link] shares several key observations about why we need this public engagement and the resulting skills gap:
This is not a problem unique to Australia.
Across 24 countries in 2026, the OECD found three in four residents five felt the political system does not allow people like them to have a say in what government does. The mechanisms for consultation have barely changed in a generation - public meetings, written submissions, comment forms. They are all designed for a slower, smaller-scale, and less interconnected public. The latest 2026 OECD Trust in Public Institutions and June 2026 Australian Public Service Commission Trust and Satisfaction in Australian Democracy Survey confirms a growing demand for more meaningful engagement, alongside a deepening sense that voices are not being heard.
The demand is real, but the supply is not ready.
A 2019 The GovLab, NYU, Monash and ANZSOG survey of nearly 700 Australian and New Zealand public servants found that engagement was among the least practised innovation skills in the public sector despite 73 percent of respondents saying they wanted to learn it. Those who received formal training were nearly three times as likely to use the skill. The systems to deliver that training, however, largely did not exist. That gap has only become more consequential as AI reshapes what is possible.
AI and new technologies, when designed responsibly in the public interest, make new forms of engagement possible.
The paper draws on ten practical examples to show what is already possible. In Bogotá in 2024, a WhatsApp chatbot called Chatico guided 50,000 residents through participatory budgeting in thirteen days with 57 percent women participants and strong representation from migrants and people with disabilities, groups that rarely appear at traditional public meetings. Brazil’s Senate is experimenting with AI to interpret over 1,000 monthly public proposals through its e-Citizenship Portal, identifying which ideas align with bills being drafted. Mexico’s Supreme Court deployed AI to analyse contributions from national justice reform forums across 31 regions. In South Australia, a citizens’ jury on cycling safety produced 21 recommendations, 18 of which the government implemented, reducing serious cycling casualties by 28 percent. The risks are equally real: AI tools that misread tone, miss cultural nuance, or produce summaries that smooth out the diversity of view that engagement is designed to surface.
The skills gap is solvable. Skills to design and use AI as a tool matter for practitioners, however, a course alone is not a workforce strategy. Evidence suggests three conditions are needed together: skills to design and run meaningful engagement; digital infrastructure matched to the task; and leadership that treats public input as a meaningful and routine part of policy design rather than an optional extra.
Where to Start
Meeting this demand requires governments that know how to listen, deliberate, and act on what they hear. What follows are examples of governments already putting this into practice.
Listening at scale, responding in meaningful ways.
The paper shows how multiple government jurisdictions are learning to use technology to listen and respond locally, while also engaging at a scale and pace that was not previously possible. In Hamburg, making sense of citizen feedback from a single engagement once occupied five full-time employees for more than a week, limiting capacity for even larger conversations. Now, the city’s Digital Participation System (DIPAS) integrates AI into the platform so that the city can cluster thousands of comments by theme, distinguish problems from proposals, and feed structured summaries directly into planning decisions.
Building skills for public professionals.
The paper also sets out how to build those skills across the public service. A new free self-paced online-course called, Designing Democratic Engagement with AI: Practical Strategies for Public Participation offers any public professional a practical framework for planning and running public engagement processes that produce meaningful outcomes, with AI as a capable assistant rather than a substitute for human judgment. The course introduces a nine-step framework, illustrated with cases from more than twenty countries, covering everything from defining goals and identifying participants to reviewing contributions and closing the loop so communities can see how their input shaped decisions.
Taken together:
Governments have an opportunity to invest in the capability to do public engagement well by teaching engagement skills.
Governments also have an opportunity to invest in designing, building and using digital platforms for public engagement.
Governments have an opportunity to make engagement routine practice for how they govern and deliver.
The discussion paper builds on two earlier ARDN papers: ARDN Discussion Paper 12 on Australian perceptions of AI, ARDN Discussion Paper 10 on the importance of getting feedback process right before introducing a new assistive technology.
Beth Noveck will be holding public talks over the next month, drawing on her new book Reboot: AI and the Race to Save Democracy:
Canberra | Reboot: AI and the Race to Save Democracy, ANU, 10 August
Canberra | APS Digital Professions Network In Conversation with Professor Noveck, Online, 13th August (Open to all APS employees)
Adelaide | Multiple events at the Democracy Innovation Festival, 17–18 August
Sydney | Can AI Save Democracy? In Conversation with Professor Beth Noveck, UTS, 27 August
I wonder if similar thinking could be applied to learning within government systems? Having preferences and opinions heard is one thing for those that wish to engage that way. Could we also listen and learn from the millions of data points already in the system… choices, requests, complaints, case notes… which might add the richness of people's lived experience and enacted preferences, allowing us to amend and create policy based on real world applied learning?
Thank you for this publication. The ability to listen at much greater scale raises a group-dynamics problem for me: crowd wisdom depends not only on the number of voices, but also on preserving sufficiently independent information before it is aggregated.
If AI clusters, labels and summarises thousands of contributions before decision-makers see them, it is no longer only reducing noise; it is partly determining which differences remain visible.
What evidence would tell us that AI-assisted engagement is improving collective intelligence rather than producing more legible consensus?
And how would you test whether the signal being compressed away contains exactly the outlier information that the majority—and perhaps the model—does not yet understand?