Leveraging complex systems: Leading for transformative change

Synthesis: Dawson and Pardo (2026) argue that generative AI (GenAI) is precipitating a systemic, paradigmatic transformation of education — not a passing fad — and that traditional bureaucratic leadership structures are ill-suited for its pace, scale, and sociotechnical nature. They introduce the SPARK framework (Systems, Problem, Analysis, Research, and Knowledge brokerage), a pragmatic model that operationalizes Complexity Leadership Theory (CLT) to help educational leaders navigate the tension between institutional stability and systemic innovation, translating GenAI-enhanced pedagogy from isolated pilots into scalable, institutionally embedded practices.

Key Findings

• Education is understood as a complex adaptive system undergoing simultaneous epistemological and structural instability; GenAI destabilises foundational assumptions about knowledge production, learner agency, assessment validity, and the role of educators as epistemic authorities.
• The paper layers three theoretical lenses: Kuhn's paradigm shifts (why education faces epistemic rupture), Carlota Perez's techno-economic transitions (how systemic changes unfold over time, with institutional lag creating turbulence), and Uhl-Bien's Complexity Leadership Theory (how leaders navigate instability by balancing bureaucratic stability with emergent innovation).
• CLT distinguishes three leadership functions — administrative (stability, efficiency), adaptive (fostering innovation in response to complexity), and enabling (creating conditions for the other two to interact productively in "adaptive space"). The critical leadership task is judging when to reinforce administrative stability and when to enable adaptive space, maintaining "productive disequilibrium."
• The SPARK framework comprises five interdependent components: Systems (mapping formal/informal structures, power dynamics, and leverage points), Problem (diagnosing adaptive challenges rather than surface technical problems), Analysis (data collection and recursive sense-making to surface hidden patterns), Research (integrating empirical/theoretical knowledge as an evidence base and feedback mechanism), and Knowledge Brokerage (activating and connecting networks of influence to scale innovation from pilots to systemic change).
• Drawing on Centola and Macy's "complex contagions," the paper argues that high-risk innovations like GenAI require multiple, reinforcing exposures from trusted social networks (faculty learning communities, cross-functional teams, student-led initiatives) to be adopted — change diffuses through trust-based, socially embedded networks, not linear knowledge flow.

Study Design & Method

This is a theoretical/conceptual paper. Drawing on Kuhn's theory of paradigm shifts, Perez's techno-economic framework, Uhl-Bien's Complexity Leadership Theory, and related work (Relational Leadership Theory, adaptive leadership, Meadows' leverage points, Centola & Macy's complex contagions), the authors develop the SPARK framework as a pragmatic operationalization of CLT for the AI era. Each SPARK component is mapped onto the CLT leadership functions (e.g., Systems→administrative, Problem→adaptive, Knowledge Brokerage→enabling). The paper is grounded in prior empirical work on learning analytics adoption (Colvin et al., 2015; Dawson et al., 2018) demonstrating that technical success alone is insufficient for institutional change. No primary data were collected.

What this means for practice

• Administrators. Diagnose GenAI challenges as adaptive rather than technical, and socialize a precise, contextual problem statement; a goal such as "increase GenAI usage" is too vague to guide institutional action.
• Administrators. Map the system before acting: stocktake the policies, processes, platforms, and stakeholders involved, and identify the governance structures the new practice requires but the institution does not yet have.
• Administrators. Broaden analytics beyond tool uptake to student–AI interaction traces, authorship and co-production patterns, AI-mediated feedback, bias and fairness concerns, and any shift in assessment validity.
• Administrators. Broker knowledge through dense, overlapping networks — faculty learning communities, cross-functional teams, student-led initiatives — because high-risk innovations need multiple reinforcing exposures from trusted networks rather than a linear rollout, and distribute agency across staff and students.
• Instructors. Treat pedagogy as the core of the transformation: redesign what counts as evidence of learning and how feedback is generated rather than relying on detection systems, and approach algorithmic personalization critically, since it risks narrowing educational purposes and displacing the relational and collective dimensions of learning.

Limitations

As a conceptual paper, SPARK is a proposed framework that has not been empirically validated; the authors present it as a practice-oriented toolkit rather than a tested model. It draws on examples from prior learning-analytics adoption research (mainly Australian higher education) rather than new data on GenAI adoption specifically. The framework's abstractions (complex contagions, leverage points, adaptive space) require contextual translation by leaders, and the paper does not provide detailed case studies of SPARK in use. The focus is on higher education leadership; applicability to other sectors (K-12, vocational) is not detailed.

Citation

Dawson, S., & Pardo, A. (2026). Leveraging complex systems: Leading for transformative change. Computers and Education: Artificial Intelligence.