Learning Framework

The SPARK learning framework translates complexity-informed ideas, such as iterative experimentation, continuous learning and an emphasis on direction rather than prediction, into an operational method.

Making Learning Operational in Collaborative Research and Innovation Systems: The SPARK Learning Framework

Over the past decade, linear models of planning and evaluation have increasingly given way to approaches that recognise complexity, uncertainty and emergence. Frameworks such as Human Learning Systems, the Cynefin model and Vector Theory of Change have contributed to this shift by offering ways to think differently about how organisations operate in unpredictable environments. Yet translating these ideas into everyday practice remains difficult.

This became clear as we developed a learning approach at SPARK, Cardiff University’s Social Science Research Park. SPARK brings together research centres, public bodies, businesses and third sector organisations to address epistemic injustice and improve lives. It is also an experiment in how to innovate social science research – testing new ways for researchers and partners to work together, generate insight and influence systems. Much of its work is collaborative and emergent: partnerships form unpredictably, projects evolve, and outcomes often appear in unexpected places.

Traditional evaluation is poorly suited to this environment, yet abandoning structure isn’t helpful either. Learning needs to be systematic enough to support decisions, while remaining light enough for a small operational team.

The SPARK Learning Framework emerged from this challenge. It translates complexity-informed ideas – such as iterative experimentation, continuous learning and an emphasis on direction rather than prediction – into an operational method. The real challenge was turning these kinds of ideas into a routine practice for learning from ongoing activity.

From data to learning signals

Organisations already generate a large amount of information through events, partnerships, surveys and conversations. Instead of treating these as isolated evidence, we treat them as learning points. Over time, these accumulate into signals: some about activity within SPARK, others about influence beyond it. Some are small interactions; others reflect shifts in relationships or ways of working.

For example, activities such as the Method Magpie programme or community networking events often begin as one-off interactions, but repeated signals – participants applying methods in their own work, forming new collaborations, or returning to subsequent sessions – accumulate into evidence of bigger change. When mapped over time, these small learning points start to cluster, indicating a shift from isolated engagement towards more sustained, cross-sector ways of working.

Learning through patterns and direction

Mapping these observations helps reveal patterns that might otherwise remain invisible. Rather than asking whether a predetermined outcome has been achieved, the more useful question becomes: what direction is SPARK moving in? This opens new conversations. Instead of debating whether something has “worked”, we can explore which adjacent possibilities look worth pursuing next to achieve our aims. Those possibilities then feed back into planning, experimentation and collaboration.

A framework for continuous learning

In this sense, the framework is less about evaluation and more about structured collective learning. Insights are gathered continuously, interpreted during regular sense‑making sessions, and translated into practical adjustments or experiments. Over time, this creates a learning cycle that helps SPARK stay oriented in an unpredictable environment.

Organisations such as the Cynefin Company have applied complexity-informed methods across several sectors, including public policy, healthcare, crisis response and organisational strategy. What we have tried to do at SPARK is bring several ideas together into a method that is light enough to sustain without large evaluation exercises or significant additional resources, and that would work in a pioneering university research and innovation setting. It continues to evolve as we learn more.

Why this matters for universities

Universities increasingly work in spaces where change is systemic rather than linear. Research partnerships, civic engagement, innovation ecosystems and interdisciplinary collaboration all involve multiple organisations interacting over time, producing outcomes that rarely stem from a single project. Yet evaluation approaches often remain tied to discrete programmes or individual outputs. If universities want to play a stronger role in addressing societal challenges, they will need better ways of learning about the systems they inhabit - approaches that capture collaboration, influence and emergent change, complementing traditional measures of research activity.

The SPARK Learning Framework is one attempt to move in that direction.

An open approach

Our learning framework is designed not to be a finished model but also an evolving practice. If it helps others working in complex systems shift from talking about learning to embedding it, that would be a valuable contribution. We would be very interested to hear how others are approaching similar challenges.