Simulating a Teenager Is Not Consulting One
Artificial Intelligence is advancing at a rapid pace, and businesses and governments are increasingly using AI tools in planning, decision-making and communicating with stakeholders. For policy actors under time and cost pressure to deliver results, AI simulations to test interventions and predict outcomes can be useful. But if this is the future of policymaking, what are the risks and opportunities? In today’s post, Srithan Gade - a senior at Lenape High School in New Jersey and co-founder of VenoxAI, a technology services business - explores this question in light of an article recently published on this blog about the importance of listening to young people whose lives are impacted by policy decisions. Srithan argues that the boundaries between simulations and consultations must always be clear and documented.
The Behavioural Insights Team - a global research and innovation consultancy - reports modelling 15 stakeholders as AI agents for a UK government department that needed to anticipate objections within weeks. Officials tested three policy approaches in a half-day workshop before real consultations began. [1]
The potential risk of this approach, under time and cost pressure, is treating simulated reactions as evidence that people have been heard.
Consider two fictional preparations for a council workshop on youth services. In the first, adult staff ask a chatbot playing a fictional 15-year-old to answer potential questions. It recommends shorter sessions and less formal language. Staff modify their plan.
This could be useful practice, but no young person participated. The simulation cannot establish that a teenager was informed, expressed a view or had an opportunity to influence the decision.
In their September blog post “‘No one wants to be the odd one’: Growing up as a teenage boy today”, Aiden and Zach, writing with Sharon Bessell and Cadhla O’Sullivan, ask adults to understand young people’s circumstances before making assumptions. Their contribution provides a starting point: ask the people whose lives you discuss. [2]
UNICEF’s 9 October 2026 article makes the principle clear: participation involves children having agency and opportunities to shape decisions affecting them. A simulated persona, by itself, gives no actual child an opportunity to participate. [3]
Even a useful prediction is a prediction
The issue does not depend on proving that AI would misinterpret a teenager’s position. Even if a model correctly predicts a concern, the simulation itself gives no young person a chance to present it, explain its significance or challenge an adult’s interpretation.
Adult staff members can use predicted positions to question their assumptions. These should be treated as ideas for further checking, not consultation results. The problem arises if “the model suggested” turns into “young people told us”.
This also concerns who sets the questions. A model’s answer, by itself, cannot establish that young people find the topic important. Additional fictional profiles would not change this lack of participation.
Provide young people with an opportunity to change something
In the second fictional scenario, staff ask young people to help prepare the workshop before the agenda is formed. They explain what decisions are open to discussion and what constraints exist. Participants are invited to challenge assumptions behind the questions, add missing topics or say that a proposed format would not suit them. For example, participants could challenge the assumption that poor attendance reflects a lack of information about the workshop and ask to discuss transport costs.
Ireland’s participation framework, developed with Laura Lundy, provides a useful check: a safe, inclusive space to form views, support to communicate them, an audience that listens, and views taken seriously and acted on where appropriate. The presence of young people in the room is not enough. [4]
In this case, the responsible staff member reviews participants’ suggestions and modifies the agenda where appropriate. They explain to participants what changed and why some suggestions were not adopted. Influence does not involve accepting every request. It means offering participants an opportunity to influence the decision and explaining how their views were taken into account.
Participation should be voluntary, accessible and safe. Organisers need to make space for people with varied experiences and communication preferences. Inviting a small group does not establish the views of all young people. [5]
Preserve the truth in records
The task is to maintain this boundary throughout the consultation process. Preparation notes and the final report should indicate the sources of ideas: simulated rehearsal, existing evidence or participants’ contributions. Simulated answers should never be counted as additional young people consulted.
Participants should have an opportunity to check the accuracy of summaries of their views, including disagreements. If some groups were not contacted, organisers should report that gap and plan how to address it. Creating a simulated persona instead of reaching out to an absent group cannot establish that its members were heard.
This boundary concerns adults inventing respondents. Young people can help create personas. [6] They should also be able to choose tools that support their communication. Participation in creating personas does not mean endorsing all future answers of the simulated respondent. New answers should not be attributed to participants without their review and voluntary agreement.
Organisations should devote enough money and time to recruitment, participation and feedback. If using AI reduces preparation time, they should use the time saved to support these activities. It should not become a reason to remove the opportunity to participate.
A useful record should protect participants’ privacy and indicate who participated, what they could influence, who considered their views and how those views were used. A convincing simulated conversation cannot establish that this happened.
Before calling a process youth consultation, adults must be able to explain how young people participated in it.
Sources
[1] Behavioural Insights Team. “AI strategic simulation.” Undated case study. Accessed 27 October 2026. https://www.bi.team/focus-areas/ai-technology/ai-strategic-simulation/
[2] Aiden and Zach with Sharon Bessell and Cadhla O’Sullivan. “‘No one wants to be the odd one’: Growing up as a teenage boy today.” Power to Persuade, 25 September 2026. https://www.powertopersuade.org.au/blog/no-one-wants-to-be-the-odd-one-growing-up-as-a-teenage-boy-today/25/9/2026
[3] Didem Özkul and Steven Vosloo. “Why children’s perspectives belong at the centre of AI policy and design.” UNICEF Innocenti, 9 October 2026. https://www.unicef.org/innocenti/stories/why-childrens-perspectives-belong-centre-ai-policy-and-design
[4] Department of Children, Equality, Disability, Integration and Youth, Ireland. National Framework for Children and Young People’s Participation in Decision-making. 2021. Relevant printed pages: 7, 11, 14–15, 18–19 and 24. https://hubnanog.ie/wp-content/uploads/2023/06/5587-Child-Participation-Framework_report_LR_FINAL_Rev.pdf
[5] UNICEF. Engaged and Heard! Tip Sheets. Written by Claire O’Kane. Guidance on meaningful adolescent and youth participation. https://www.unicef.org/youthledaction/documents/engaged-and-heard-tip-sheets
[6] Kathleen W. Guan and colleagues. “Leveraging Participatory Personas for Reflexive Co-design of Personalization in Large Language Models.” UMAP 2026, pp. 506–509. DOI: 10.1145/3774935.3812729. https://research.tudelft.nl/en/publications/leveraging-participatory-personas-for-reflexive-co-design-of-pers/
Author bio and competing interest
Srithan Gade is a senior at Lenape High School in New Jersey and co-founder of VenoxAI, a technology services business. He has a commercial interest in technology implementation and support through that business.
AI use disclosure
ChatGPT helped with research and formatting. I wrote this version and used ChatGPT for grammar edits, factual and source checks, and document formatting.
Content moderator: Sue Olney