Epistemic Defragmentation: Connecting Diverse Forms of Knowing in Complex Systems

Contemporary challenges in earth system science, such as climate change, disaster risk and emerging technologies are characterised by complexity, uncertainty and multiple, often incomplete representations of reality. Yet the knowledge required to understand and respond to such challenges is increasingly fragmented across disciplines, methodologies, institutions and epistemic communities. Scientific models produce formal representations; practitioners contribute experiential and situated knowledge; communities hold contextual and place-based understandings; policy institutions require actionable knowledge; and artificial intelligence increasingly produces synthetic forms of knowledge by aggregating and recombining information. These forms of inquiry are not simply different sources of information: they embody different ways of framing problems, establishing evidence and determining what counts as knowledge. The challenge is therefore not merely to accumulate more knowledge, but to understand how heterogeneous forms of knowing can relate when none provides a complete representation of a complex system.

This paper develops epistemic defragmentation as a philosophical response to this problem. Building on several bodies of philosophical scholarships, it asks: how can diverse and necessarily partial forms of inquiry be brought into productive relationship without reducing their epistemic differences? It argues that epistemic unity should not be understood as the homogenisation, aggregation or synthesis of knowledge into a singular representation. Rather, unity can be understood relationally: as the capacity of different forms of inquiry to become mutually intelligible while retaining their distinctive assumptions, purposes, methods and limitations. Epistemic defragmentation, therefore concerns not the elimination of epistemic plurality, but the creation of relationships among otherwise disconnected forms of knowing.

The paper further argues that epistemic defragmentation must be grounded in epistemic humility: the recognition that no single form of inquiry can fully capture a complex reality. Such humility requires us to acknowledge both the strengths and limitations of our own knowledge and to remain open to what other forms of inquiry may reveal. Rather than treating uncertainty, disagreement and incompleteness as deficiencies to be eliminated, epistemic humility creates the conditions for different forms of knowledge to question, complement and refine one another. Epistemic defragmentation thus provides a way of thinking about unity without requiring uniformity.

Drawing on complex Earth-system science, where atmospheric, ecological, geological and social processes interact across spatial and temporal scales, the paper illustrates why such an approach is necessary. Climate and disaster management knowledge is produced through models, observations, historical records, expert judgement, local experience and increasingly computational and AI-mediated forms of analysis, each offering only a partial representation of dynamic systems. In such contexts, uncertainty and disagreement cannot always be resolved through additional information or more sophisticated modelling. What becomes necessary is the capacity to relate different representations, recognise their limits and construct understanding across epistemic boundaries. Thus, this paper proposes epistemic defragmentation as a philosophical orientation for understanding complex systems rather than a barrier to it.