Values in Motion

This summer, public discussion about Artificial Intelligence (AI) has increasingly portrayed AI as a means of increasing productivity, efficiency, and cost savings in public services. As Finland faces fiscal, demographic and structural issues, AI is sometimes referred to as a “lifeline” for maintaining quality and accessibility of public services while streamlining public organizations. While some have raised concerns about job displacement, faceless AI-driven public services, and the erosion of democratic foundations, the debate often sidesteps more fundamental questions: What kind of society are we trying to build with AI? And which values are we willing to uphold, even when doing so comes at a cost?

These abstract questions become easier to grasp when considered more concretely: Would you let an AI agent vote on your behalf if it knew your political values? Would increasing use of AI in public services and organizations increase or decrease your trust in public institutions? Would you rather interact with a politician or their digital twin during elections? Would you agree to implant a monitoring chip in your child if it improved their care?

To explore these value tensions, we organized a session titled “Values in Motion” at the Wasa Future Festival 2026 (WFF 2026). Our session was grounded in two ongoing research projects, funded by the Strategic Research Council (SRC), that examine the values shaping the use of AI in society. AIdemoc explores the future of democracy in the age of AI, while Carefuture examines the future of care work in an AI-enabled society. Together, these projects share a common concern: as AI becomes increasingly embedded in society, which and whose values will guide its development and use?

Rather than presenting research findings in a traditional format, we organised the session as an interactive campfire discussion. Participants were invited to respond to provocative questions, vote on alternative futures, and identify the values they considered most important in shaping the role of AI in society. The discussion focused on three areas where AI is already influencing everyday life: democracy, public administration, and care work. Across all three, the same question repeatedly surfaced: when values come into conflict, which ones should guide the development and use of AI?

The Future of Democracy in the Age of AI

A central theme of the AIdemoc project is how AI might reshape democratic institutions and citizen participation in the future. We have developed four alternative scenarios that capture different ways in which AI might transform democracy:

AI strengthens the current democracy. Incremental improvements enable more informed decision-making and broader citizen participation.
AI transforms democracy into something radically new. Knowledge processes become more transparent, AI literacy becomes more widespread, and new forms of participation and deliberation emerge through AI-assisted systems. AI enables new democratic innovations to emerge.
AI weakens and narrows democracy. Manipulated information proliferates, votes drift to AI agents, and inequalities widen between those with access to advanced AI tools and those without.
AI collapses democracy. AI-generated information can no longer be trusted, platform manipulation becomes routine, and power shifts from citizens to large technology companies and AI developers. Collapse of trust between the public, institutions and representatives.

When we asked the WFF 2026 audience which scenario seemed most likely, the majority chose the third: AI weakens and narrows democracy.

Interestingly, few participants believed that AI would transform democracy into something radically new. A recurring argument was that democracy is such an old and institutionalized way of organizing societies that radical change appears unlikely. Democracy was seen not only as a mechanism for collective decision-making, but also as a value worth preserving in its own right.

Although participants viewed the weakening and narrowing of democracy as the most likely future, such a development is not inevitable. AI is nevertheless already becoming a visible part of democratic processes around the world, meaning that its relationship with democracy is no longer a matter of speculation alone.

Argentina's 2023 presidential election was described as the first AI election, as both leading candidates produced campaign material using AI. During the same year, three candidates in the U.S. presidential primaries deployed AI-generated chatbot versions of themselves. In 2024, Pakistan’s former prime minister Imran Khan campaigned from prison using a voice clone. The same year in the U.S., Jason Palmer defeated Joe Biden in the American Samoan primary, even though he was virtually unknown, by utilizing AI-generated outreach and an avatar that answered voters' questions on his website.

These examples do not, in themselves, indicate that democracy is being weakened. Rather, they illustrate how AI is becoming embedded in political communication, election campaigns, and interactions between politicians and citizens. Whether these developments ultimately strengthen, transform, weaken, or contribute to the decline of democracy depends on the values, institutions, and governance mechanisms that shape their design and use.

Efficiency, Sovereignty, and Public Administration

Finland’s public administration is actively promoting the use of AI to improve productivity, efficiency, and service delivery. In workshops conducted in 2024 as part of our research on AI experiments in legislative drafting, civil servants identified several tensions emerging within public administration itself. These tensions show that adopting AI is not simply a technical decision but requires public organizations to decide which values should guide development when important goals pull in different directions.

Quality or efficiency? AI may allow public organizations to process information and complete tasks more quickly. However, if efficiency becomes the primary measure of success, the quality of public services and decision-making may suffer. The important question is whether AI genuinely improves public administration or merely enables work to be completed faster and with fewer resources.

Stimulating or replacing human thinking? AI can be used to challenge assumptions, generate alternatives, and support critical reflection. In practice, however, automation may also lead to cognitive offloading, reduced oversight, and excessive trust in machine-generated outputs. The question is whether AI will strengthen the judgement of public officials or gradually reduce the amount of independent thinking involved in their work.

The best available technology or growing dependency? The most capable AI models may offer immediate benefits, but many are developed and controlled by international technology companies. Relying on these systems may make public administration increasingly dependent on technologies, providers, and development choices outside Finland’s control. Over time, such reliance could also weaken public institutions’ ability to carry out essential tasks without external systems.

Greater trust or growing distrust? The effects of AI on public trust are uncertain. AI could make administrative criteria more consistent and transparent, helping citizens understand how decisions are made. At the same time, the use of opaque models could make public decision-making more difficult to explain, challenge, or oversee. Whether AI strengthens or undermines trust will therefore depend on how transparently and accountably it is introduced and used.

These tensions cannot be addressed by prioritizing one value while disregarding others. Public administration needs efficiency, but also quality; advanced technology, but also sufficient autonomy; intelligent tools, but also human judgement. The choices made when balancing these values will help determine not only how well public AI systems function, but also how they affect democratic accountability, fundamental rights, and citizens’ trust in public institutions.

The WFF audience was particularly hesitant about public organizations using powerful AI models if doing so increased Finland’s dependence on foreign technology companies or introduced values that might conflict with those of Finnish society. At the same time, they recognized the lack of alternatives: do we have powerful enough Finnish, European or open-source AI models that meet the needs of the public administration? And could avoiding the most advanced models mean missing opportunities to improve public services or strengthen democracy? There may be no risk-free path: adopting the most advanced models can create dependency, while rejecting them may limit innovation and opportunities to use AI in support of democracy.

This raises a further question: whose values are embedded in the AI models used by public institutions?Large language models are not value neutral. The Economist recently compared the responses of major AI models with data from the World Values Survey. While most models leaned toward secular and self-expression values, their value profiles did not clearly match those of any particular country or society. If public administration wants to use models that reflect the values it is expected to uphold, someone must decide which values matter, in which contexts, and through what process. Yet there is currently no clear democratic procedure for making these choices. Developing such a procedure will be essential if decisions about public-sector AI are to remain democratically legitimate and accountable.

Where does care end and surveillance begin?

A central theme of Carefuture project is to explore how AI solutions can support care work in an ethically and socially sustainable way. Care work offers a concrete example of how AI can bring values into conflict. The growing use of AI and digital technologies in care makes it increasingly difficult to balance values such as efficiency, safety, autonomy, privacy, and human dignity. For example, many Finnish and European research and development projects are studying sensor-based technologies that monitor sleep, vital signs, functional capacity, and daily routines. These technologies may alert users and care providers to changes and enable earlier intervention.

However, patients, healthcare organizations, social and health authorities, and society at large may each interpret the benefits of these technologies in different ways. A study of chatbots in public services found that designers understood efficiency in terms of reducing service-delivery costs, whereas users associated it with receiving timely assistance and achieving the desired service outcome without unnecessary waiting (Makasi et al. 2022).

Where does care end and surveillance begin? Data collected by sensors may support earlier and better care, but it also raises questions about where the data is stored, who controls it, and how it may be used. A wristband can be removed, but an implant is much harder to refuse or reverse. Which groups will be among the first to be asked to adopt such technologies, and for whom will their use remain a genuine choice rather than a condition for receiving services?

These questions become particularly difficult when monitoring concerns people who cannot make such decisions independently, such as vulnerable elderly people with dementia, or children. For example, would an implant be justified if it helped manage a child’s medical condition, even if a decision made by a parent could affect the child’s privacy far into adulthood? The WFF discussion highlighted the need to address these concerns when technologies are designed, while also considering the inequalities that may arise if people who reject monitoring receive poorer or more limited care.

Keeping values in motion?

With the adoption of AI, our societal values are likely undergoing a process of change: gradually, new values are emerging, and old values that we hold dear are being eroded or are struggling to find application in the changing socio-technical assemblage. Efficiency may conflict with quality, technological capability with autonomy, and safety with privacy. These tensions cannot be resolved by technology alone.

We need an ongoing public discussion about the values we want to uphold, which values AI should support, how competing values should be balanced in a just process, and who gets to decide. Keeping values in motion means making these choices visible and revisiting them as technology and society evolve. We must therefore establish democratic ways to choose and balance the values guiding AI before these systems take on greater roles in decisions affecting people’s lives.

If the values of decision-makers, service providers, and service users pull in different directions, society risks losing its course. We must therefore correct our course in time and ensure that AI is designed and used to support democratically chosen values. The discussion about values in relation to the use of AI must be carried out at all levels of society and in all contexts. Only then will we continue to navigate towards a future in which it supports democracy, trustworthy public services, and human dignity.

References

Khalid, A. (2023, December 18). Pakistan’s former prime minister is using an AI voice clone to campaign from prison. The Verge.

Makasi, T., Nili, A., Desouza, K. C., & Tate, M. (2022). Public service values and chatbots in the public sector: Reconciling designer efforts and user expectations. Proceedings of the 55th Hawaii International Conference on System Sciences, 2334–2343.

Nicas, J., & Cholakian Herrera, L. (2023, November 15). Is Argentina the first A.I. election? The New York Times.

Schneier, B., & Sanders, N. E. (2025). Rewiring democracy: How AI will transform our politics, government, and citizenship. The MIT Press.

Surprise Democratic primary winner credits AI for beating Biden. (2024). Mashable.

The Economist. (2026, June 25). AI models’ values are very different from most people’s.

World Values Survey Association. (n.d.). World Values Survey database.

Nina Wessberg, Jaana Leikas, Antero Karvonen, Anton Sigfrids, and Hanna Saari. More at aidemoc.fi and ttl.fi/hoivatyon-tulevaisuus-carefuture.