A New Approach to Experiential Learning: Scholars inform the City of Helsinki on ethical AI applications for schools during BKC’s AI Policy Research Clinic

As part of the Ethics of Digitalization project, a joint initiative of the Global Network of Internet and Society Research Centers (NoC), the Berkman Klein Center for Internet & Society collaborated with the City of Helsinki’s Education Division and their subcontractor, Saidot, to host an intensive, virtual AI Policy Research Clinic. This three-week educational program invited a dozen graduate students to engage in a real-world project and apply their backgrounds in law, public policy, computer science, education, and other disciplines towards the responsible use of AI, for an application designed for Helsinki schools. The specific application aims to improve personalized learning programs, student well-being, and retention within Helsinki’s vocational education and training programs.

“The AI policy research clinic connects burgeoning experts in academia with a public-sector partner in order to develop viable approaches to a public policy challenge,” says Adam Nagy, one of the BKC program coordinators who assisted with the planning and execution of the program. “What is so unique about this model is that it demands both groups operate outside their traditional turf. The students have to generate recommendations that are constrained by the resources, timelines, and interests of a real-world example. At the same time, the partner is engaging in an educational exercise. The students used the diversity of their backgrounds and skills to generate big-picture outputs: identifying the various underlying issues, redefining the problems, mapping the ecosystem of stakeholders, and comparing different approaches.”

The early career scholars studied and contributed in two critical areas: exploring and suggesting processes for participatory design throughout the lifecycle of the technology and building human accountability structures. Nagy says, “Rather than task the cohort with trying to solve complex societal issues on their own – an approach that often misses crucial dimensions of the problem – we focused their efforts on envisioning systems and structures that will empower the people actually impacted by the technologies and policies to affect their design and governance.” By building capacity in these areas, the program organizers, participants, and partners hoped to prioritize the mitigation of potential harms – particularly in relation to issues of bias or fairness across protected categories such as race or gender – in ways that are sustainable, participatory, and transparent.

A New Approach 

The AI Policy Research Clinic, as a pedagogical approach to experiential learning and problem-solving, was itself an experiment in how might a dozen advanced graduate students rapidly learn the needs and constraints of a public-sector partner and combine these learnings with their expertise to generate potential paths forward? 

“After we’ve witnessed – and ourselves contributed to – the development of AI ethics and governance principles over the past few years, the time has come to examine how to translate these abstract norms into actual practice,” says Urs Gasser, Professor of Practice at Harvard Law School and co-lead of the Network of Centers. “Focusing on real-world use cases, this innovative program offers an opportunity to explore some of these hard translational questions in an educational environment, in the spirit of experiential learning and with a strong commitment to the public interest.”  

Helsinki’s firm ethical guidelines prohibit the categorization of students by metadata collected from or about them. No raw or anonymized student data, past or present, was provided to the research cohort. To generate viable recommendations the research clinic cohort had to interpret and apply existing regional, national, and local regulations and principles regarding the use of AI in educational settings. 

Pasi Silander, Head of ICT Development Programs at the Helsinki Education Division; Pasi Rautio, Project Manager at the City of Helsinki and RPA development; and Meeri Haataja, CEO and Co-Founder of Saidot, guided and responded to the cohort’s efforts. Saidot is a start-up with a mission for enabling responsible AI ecosystems. As Haataja describes, “It was fascinating to see how quickly students found ways to mobilize the unique expertise and experience profiles of the multidisciplinary team and apply them. While it is clear how important participation of different affected groups – students, families, teachers – is in learning analytics development, this process gave concrete answers on how to do it.”

Outcomes and Deliverables

The students offered the City of Helsinki new methodologies and frameworks to approach participatory design and human oversight. These outputs also serve as useful signposts for other municipalities seeking to use AI in their own educational contexts or in the provision of different public services. Any local policymaker considering the integration of AI might consider: how, from a participatory angle, can their city ensure that all pertinent voices are being heard? How can trust be built with new AI systems in settings as sensitive and ubiquitous as public classrooms, housing, transportation, law enforcement, and other services?

The first of two working groups created three resources: a human oversight model designed to fit into the existing governance structure of Helsinki and enhance cooperation across technical, operation, and governance organizational layers; a translational matrix based on ethical and regulatory requirements in the European Union; and a wireframe with supporting documentation for an accountability web-portal to enable public participation, enhance transparency, and support the individuals tasked with human oversight of the learning analytics and AI tools. The group consolidated their background research, justification for each resource, and implementation recommendations into a policy playbook.

The second group adapted an existing method deployed in Catalonia by Coboi Lab to the City of Helsinki’s requirements and the specifics of the AI technology. In an extensive playbook, the students describe the essential, recommended, and contingent elements of a four-phase participatory process for the introduction of the AI technologies. Policymakers using this resource will find a richly detailed, step-by-step guide to integrating participatory and accountability elements into their design, development, and deployment process.

The first group’s efforts drew heavily from AI-specific ethical guidelines and practices, while the second group adapted a more holistic approach to a use case involving AI. Ultimately, both approaches emphasized that the deployment of AI in municipal services should be a continuous exercise, with constant input from stakeholders, technical teams, and the general public as new challenges and concerns arise. These systems support a sustainable ecosystem of responsive and empowered stakeholders.

What We Learned

“Developing ethical AI – especially in municipal settings – can face a tradeoff between participatory design and swift implementation,” reflects Franziska Poszler, a research associate and doctoral candidate at the Technical University of Munich. “For example: if AI is deployed in the education system, citizens are mandated to engage with it in order to engage with public education. It becomes ever more important to include affected individuals in the design of this new technology.”

The research clinic teams addressed challenges of connecting theoretical principles to human action within material constraints. “Early career scholars who participate in programs like these get a chance not only to share what they know but also to experience the practical challenges of real world design – knowledge and insights that they can then bring back to their studies and to the field. All parties involved can benefit from this innovative pedagogical approach,” says BKC Managing Director, Lis Sylvan. 

The research clinic offers insights and lessons, not only for how municipalities approach integrating AI, but for how graduate students can further the ideas of democratizing digitization imparted by their mentors. “We learned about Helsinki’s innovative efforts and strategies of building participatory designs and education technology,” says Poszler. “Every participant of the Research Clinic can now take all of these valuable insights back to their home country or home university and try to push for similar innovations.”