The LiU Design department has been running a seminar series on mission-oriented innovation. I've been attending, trying to connect what I'm learning about federated learning to broader discussions about how innovation happens - and doesn't happen - in public sector contexts.
This post gathers my reflections on that connection. It's early in my fieldwork, and I'm still orienting. But the missions literature is surfacing questions I don't yet know how to answer.
What Are Missions?
Mazzucato's framework has become influential in policy circles. The basic structure is hierarchical: grand challenges (like climate change, or healthy ageing) are addressed through missions - concrete, time-bound, measurable goals that galvanise action. Missions are pursued through portfolios of projects, which combine to produce systemic change.
Miedzinski, Mazzucato and Ekins describe mission-oriented policies as "systemic public policies that draw on frontier knowledge to attain specific goals - in other words 'big science' deployed to meet 'big problems'" (Miedzinski et al., 2019, p. 5). The language of ambition is central: missions tackle challenges "that require major transformations in production and consumption patterns" (Miedzinski et al., 2019, p. 5).
The framework has obvious appeal. It suggests that with the right framing, coordination, and political will, complex problems can be addressed systematically. It positions the state not just as a market corrector but as a market shaper - actively directing innovation toward societal goals.
Vinnova, Sweden's innovation agency, has been piloting mission-oriented approaches around mobility and food systems. Dan Hill, who worked on this process, describes how the pilots ended up "addressing streets and school food" - which meant "addressing politically-complex societal challenges" quite different from the technological moonshots that originally inspired mission thinking (Hill, 2022, p. 2).
The Moonshot Problem
Here's where things get complicated.
The missions discourse draws heavily on the Apollo programme as exemplar. Kennedy's commitment to landing a man on the moon and returning him safely within a decade is presented as proof that ambitious, coordinated action can achieve remarkable things.
But Hill, drawing on Richard Nelson's earlier critique, notes a crucial distinction:
"The Apollo moonshot was clearly a technological mission primarily, and framed within the quite different, arguably simpler, dynamics of the Cold War... The real problem was that a purely scientific and technological solution could not solve such problems" (Hill, 2022, p. 2).
Nelson's point, made decades ago, was that social problems aren't amenable to the same kind of engineering that sent rockets to the moon. The moon doesn't have stakeholders. It doesn't have competing interests, legacy systems, or institutional inertia. You can't prototype your way to solving poverty in the same way you can prototype a lunar module.
Hill puts it directly: "Technological and economic solutions are good at fixing technological and economic problems" (Hill, 2022, p. 2). The challenge for mission-oriented innovation applied to social challenges - health, welfare, education - is that these involve what design theorists call "wicked problems": situations where the problem definition is itself contested, where interventions change the problem, where there's no stopping rule for when you've succeeded.
Dark Matter
Hill's earlier work introduced the concept of "dark matter" in strategic design - the organisational culture, policy environments, market mechanisms, legislation, and finance models that shape what's possible but remain largely invisible:
"The dark matter of strategic designers is organisational culture, policy environments, market mechanisms, legislation, finance models and other incentives, governance structures, tradition and habits, local culture and national identity, the habitats, situations and events that decisions are produced within. This may well be the core mass of the architecture of society, and if we want to shift the way society functions, a facility with dark matter must be part of the strategic designer’s toolkit". (Hill, 2012, p. 7).
The relationship between dark matter and visible outcomes is what makes strategic design different from traditional design practice. You can't design a transformative service without engaging with the organisational context that will produce and sustain it. Hill argues that "strategic design recognises that this 'dark matter' is part of the design challenge" (Hill, 2012, p. 7).
This resonates with something I'm starting to notice in my own fieldwork. The FL proposal I'm working on assumes certain things exist - data infrastructure, technical capacity, governance frameworks, stakeholder alignment. These are the dark matter that would need to be in place for FL to work. But I don't yet know whether they exist. The proposal was developed at some distance from the implementation context.
What Missions Require
Reading the missions literature more carefully, I'm struck by how much it presupposes.
Miedzinski et al. emphasise that "seeking stakeholder alignment is particularly important for missions with an ambition to enable transformative system innovations" (Miedzinski et al., 2019, p. 14). Missions require "working closely with stakeholders in designing and implementing mission-oriented policy... to ensure a greater buy-in, and commitments to invest in mission-related activities" (Miedzinski et al., 2019, p. 14).
The framework describes missions as addressing challenges that are "complex, multidimensional, dynamic and uncertain in the long run" (Miedzinski et al., 2019, p. 9). The response to this complexity is portfolios - "multiple efforts" that align to produce systemic change (Hill, 2022).
But what does this mean in practice, at the local level? Vinnova operates at national scale with significant resources. The coordination association I'm working with operates at municipal scale with ESF project funding. The gap between where mission thinking is developed and where it might be implemented seems significant.
Applying This to Federated Learning
Let me try to apply the missions framework to my own project.
The grand challenge might be framed as: improving outcomes for people navigating the welfare system back toward work or education. This is genuinely important - people get stuck in bureaucratic limbo, passed between agencies, their situations deteriorating.
The mission might be: enabling data-driven decision support that helps caseworkers and clients identify promising pathways. Concrete, potentially measurable, time-bounded.
Federated learning would be one project in a portfolio - a technical approach that allows organisations to collaborate on predictive models without sharing sensitive individual data.
But here's where the framework starts to strain.
The missions literature assumes you're working at the level of innovation policy - coordinating research funding, convening stakeholders, shaping markets. I'm a PhD researcher embedded in a single organisation, part of a small academic-practitioner collaboration. I don't have the positional authority to "seek stakeholder alignment" across the Swedish welfare system.
More fundamentally, FL presupposes infrastructure that I don't yet know exists. It's a solution to a specific problem: how to collaborate on machine learning when data can't be centralised. But that problem only arises if you already have data infrastructure, technical capacity, and governance frameworks at each federated node.
What if the dark matter isn't in place? What if the preconditions for the mission aren't present?
Questions I'm Carrying Forward
The seminars have been intellectually stimulating. Mission-oriented innovation offers a compelling vision of how public sector transformation might work. But I'm left with questions:
Scale mismatch: Missions are typically conceived at national or EU level. How does mission thinking translate to local implementation contexts? I'm curious whether grand frameworks help or hinder the prosaic work of building infrastructure and capacity.
Preconditions: The missions literature focuses on alignment, portfolios, and coordination. But what if the basic preconditions - data, systems, skills - don't exist? This is something I need to investigate in my fieldwork.
Technological vs social: FL is a technical solution. But the problems of vocational rehabilitation seem to be social, institutional, and political. I'm trying to understand how technical approaches fit - or whether they fit at all.
The research-practice gap: The academics who proposed FL for Swedish vocational rehabilitation developed the idea at some distance from the implementation context. How common is this pattern? How do good ideas generated in academic contexts survive the transition to implementation?
I don't have answers yet. I'm still mapping the terrain. But the missions literature is useful precisely because it highlights what's needed for ambitious innovation to succeed - and by implication, what might be missing.
In my next post, I want to think more systematically about what FL would actually require. Not the technology itself, but the pyramid of preconditions that would need to be in place before FL becomes a meaningful option.
References
Hill, D. (2012). Dark Matter and Trojan Horses: A Strategic Design Vocabulary. Strelka Press.
Hill, D. (2022). Designing Missions: Mission-Oriented Innovation in Sweden. Vinnova.
Miedzinski, M., Mazzucato, M. and Ekins, P. (2019). A framework for mission-oriented innovation policy roadmapping for the SDGs: The case of plastic-free oceans. UCL Institute for Innovation and Public Purpose Working Paper Series (IIPP WP 2019-03).