Public affairs research: policy testing in 3 phases

Public affairs research can expose the position you want to test. In the transport-policy engagement, a global transportation leader needed to choose a narrative before committing publicly. Its intended audience included lawmakers and regulators, alongside journalists and advocacy groups. Showing those people draft messages would have altered the debate under examination, according to our transport-policy case study.

Recruitment was therefore part of a wider problem. Senior policymakers and regulatory officials are hard to reach through conventional research, but securing an interview would not have resolved this client’s concern. The act of presenting an alternative policy position to an influential stakeholder was itself an exposure. The research needed to preserve the client’s freedom to test arguments it might later reject.

We used simulated stakeholders so the client could compare frames in isolation and examine adverse scenarios without putting those frames into circulation. The constraint in this case was sensitivity, not a stated lobbying-disclosure rule. The relevant result is concrete: the client could examine reactions to draft positions without a message reaching a live audience.

Public affairs research and government affairs research address overlapping decisions, with different emphases in United Kingdom and United States usage. Government affairs and government relations are more common United States terms for direct engagement with legislators and regulators. Public affairs is broader in United Kingdom and European practice, including media, public opinion and coalition work around that engagement. Treat these as conventions of usage, rather than a formal division of responsibilities; the public affairs industry overview (opens in a new tab) describes a field whose terminology varies by country.

The Chartered Institute of Public Relations used “public affairs practice” in its announcement of its lobbying conduct guide, published on 16 July 2015 (opens in a new tab). The CIPR says its principles of integrity, competence, transparency and confidentiality should apply in public affairs practice. That professional context covers the conduct of engagement; our research prepares a position for it.

For research design, specify the audience before settling the label. Our transport-policy society included advocates, journalists and industry voices as well as lawmakers and regulators. A question about a proposed policy narrative therefore extended beyond how officials would receive it. The design also examined what happened when damaging coverage entered the discussion.

For the transport-policy project, we built Washington D.C. Opinion Leaders: a society of 1,500 AI personas spanning more than 360 organisations and 800 distinct job titles, according to our transport-policy case study. Each persona was matched one-to-one to a real individual through public data and social activity. The society represented people who draft, interpret and debate national policy, including advocates and industry voices.

In Washington D.C. Opinion Leaders, the construction connected the intended policy audience to observations of real individuals, while the analysis examined differences by party, gender and role. You could inspect reactions below the aggregate result.

Our method and evaluation page describes how we ground each persona in a real individual and connect personas in a network of influence. Public observations or your own data can inform the audience definition. The method depends on those observations: where diverse observations do not exist, there is nothing to ground a persona in. That boundary belongs in the initial discussion of the audience, before you settle the narrative test.

The transport-policy research compared narratives under changing conditions. We began with baseline attitudes, blind-tested two competing narratives and their proof points, then introduced damaging coverage and hypothetical scenarios, as described in our transport-policy case study. A proof point here is the evidence offered in support of the policy argument. The design asked both whether the argument persuaded and whether its supporting evidence retained credibility under pressure.

Source: Artificial Societies, transport-policy positioning case study.

The sequence matters to the decision. A favourable response to a narrative tells you about that presentation. Examining the response after damaging coverage asks whether the recommendation remains credible when the surrounding story changes. We would want that second test before recommending a policy narrative whose reception depends on contested claims.

Because the participants were simulated, we could repeat tests and examine worst-case scenarios without asking real stakeholders to respond to those materials. We captured results at aggregate and individual persona level, with analysis by audience segment, narrative and proof point. The client could therefore examine which argument worked with which part of the audience, instead of relying on an overall preference alone.

The transport-policy simulation diagnosed an existing credibility gap and identified the strongest narrative and supporting proof points. It also showed where those arguments held up or collapsed under negative coverage. Our transport-policy case study reports 250,000 individual responses from more than 170 questions across the three phases.

We delivered a flagship advisory report with recommendations on positioning, proof points and risk. The client also received full access to the Artificial Societies platform, so its team could examine segment cuts and individual persona reactions themselves. The report gave the team a recommendation; access to the underlying reactions let it examine the basis for that recommendation when considering a particular argument or audience group.

The engagement took three weeks from start to finish, including construction of the society, according to the same case study. No message reached a live audience. That duration describes this project, rather than a delivery promise for a different audience or set of questions.

We would judge the assignment by the choice it clarified. Diagnosing a credibility gap gives a policy team something to address before it commits to a narrative. Showing where proof points withstand adverse coverage gives the team a basis for choosing what to substantiate. Those are decisions the research supports; a regulatory decision remains outside the claim.

Government affairs research can examine reactions to regulatory submissions and consultation responses, or support preparation for legislative hearings. Other applications include testing the sequence of a policy announcement and shaping coalition or stakeholder-engagement strategies. We would scope those assignments around a proposed position and the audience’s response to it. The transport-policy example directly demonstrates narrative comparison and testing under damaging coverage.

Policymakers and Regulators also formed a named audience in our crisis-simulation case study, alongside Media, Influencers, Consumers and Commercial Customers. That project examined messages after a crisis and informed recommendations on communications and delivery. For an advocacy campaign or coalition assignment, make the equivalent audience choice explicit: the Washington transport design examined differences by party, gender and role.

A simulated reaction may suggest where to revise a submission or prepare an answer for a hearing. It cannot establish how an official will exercise authority. We would use the comparison to prepare the position you take into engagement, with the recommendation based on the tested materials and the audience represented in the society.

Simulated public affairs research examines responses to policy positions, including differences between audiences and reactions to adverse coverage. Our scope is diagnostic and comparative. We do not claim to predict legislative votes or regulatory outcomes, and we would not turn a preferred narrative in the Washington study into a probability of policy success.

Grounding is the other boundary. Diverse observations give personas a basis in the people you need to understand. Where that basis is absent, adding detail to a persona’s description does not provide the missing observations. Your discussion with us should therefore begin with the audience and the evidence available about it.

Our Survey Evaluation Report, January 2026, also records 89% internal coherence, measured as Cronbach’s alpha: alignment between questions measuring the same underlying attitudes (method and evaluation). For a policy team, that concerns whether related answers fit together. The same report gives 67% distribution accuracy for biography-prompted large language models. Its benchmarks let you assess the evidence behind the method before using a narrative comparison to inform a recommendation.