This document is an institutional design treatise about theories of change and potential struggle to fully account for the complex interplay of variables and uncertainties that can lead to unforeseen consequences or vulnerabilities in institutional systems.
Introduction
An organization is a dynamic system that depends on complex variables at each point of its evolution. If we try to model an organization as an evolutionary system then its change and impact in the real world is complex and non-deterministic.
The premise is that : we have not been experimenting with new institutional structures for doing research and thus do not have a good idea on how they are different from regular organisations and theories. One good example of this is speculative tech with their coordinated research programs(they are putting together a guideline on how to do this with ARIA). Perhaps this could much more approximate mechanism design for institutions.
AFAIK, a theory of change for a Manhattan or Biosecurity Project is different from an organisation operating with a monolithic goal.
Alternatives to Theory of Change
Theories of change, by their nature, are constrained by our posterior understanding and presuppositions. They often struggle to account for the full spectrum of variables and uncertainties inherent in organizational evolution.
Most Significant Change
A participatory stakeholders who come together and collate beneficiary opinions on what matters for them
What has been the most significant change that you have seen in the intervention and how has this impacted you.
Realist Evaluation
The UK Public Health uses realistic evaluations which diverge from theory-based methods by emphasizing the interplay of Context and Mechanism to produce desired Outputs. This approach acknowledges the complexity of social interventions, recognizing that outcomes are shaped not just by the intervention itself but also by the context in which it operates and the underlying mechanisms through which change occurs. The relationships between contextual factors, mechanisms, and outcomes, thereby enabling a more nuanced analysis of program effectiveness and impact. By focusing on these dynamic interactions, Realist Evaluation offers a robust methodology for uncovering the underlying processes driving program success or failure in real-world settings.
Institutional Design
Institutions are antifragile in the short-term and may sustain durability in the long term (eg. legal systems, democracies, universities and religious institutions). However the sustenance of such institutions can evolve in ways that are unrecognizable to their founders’ original visions.
Nick Bostrom mentions in the vulnerable world thesis:
“For example, a well-intentioned project may be subverted in its
implementation; or it might turn out to have bugs or institutional
design flaws that become apparent only after a period of normal
operation. Even if the system itself functions precisely as intended
and remains un-corrupted.”
While theories of change provide a framework for understanding and managing organizational evolution, they may fall short in addressing the intricate interplay of factors that can lead to unforeseen consequences or vulnerabilities in institutional systems.
Delta of Change (Path finding)
Introduction
The "Delta of Change" refers to a theory or framework for managing change by identifying elements that can alter the direction of change and determining strategies for approaching various paths or trajectories. In this framework, the "delta" represents the potential changes or disruptions that can influence the direction of a process or outcome. This could include factors such as new information, external events, stakeholder input, or emerging trends. The approach involves assessing the potential deltas or changes and developing strategies for navigating them within the context of identified paths or trajectories. This could include contingency planning, adaptive strategies, or proactive interventions to address potential shifts in direction.
Differential Change
Technological development is mainly differential, differential technological development refers to the notion that advancements in technology do not progress uniformly across all areas or domains. Instead, development occurs differentially, with some fields or sectors experiencing rapid progress while others lag behind. One reference to explore this concept further is Nick Bostrom's paper titled "Strategic Implications of Openness in AI Development," where he discusses the idea of differential technological development within the context of artificial intelligence (AI). Bostrom explores how different AI capabilities may progress at varying rates, leading to differential impacts on society and raising important strategic considerations for AI development.
Another reference is the work of Eric Drexler, particularly his writings on nanotechnology and molecular manufacturing.
Delta of Change
Increasing organizational deltas should deform the theory of change and help create new realities for the organizations. Deltas of change are directly inspired by differential approaches to technological development, such as Responsible Scaling Policy(RSP) from Anthropic. At different AI Safety Levels as a way to measure powerful models scaling outside of safety procedures. This model is a Delta of Change because it is not monolithic but considers capabilities and safety measures hand-in-hand. This framework helps Anthropic decide checkpoints that are necessary for the deployment of Claude. This same method can be analogous to how organizations can deform and form their theories of change based on external factors.
OODA-lookalike Loop
OODA loops tend to have a differential approach with an orient-observe, decide and act framework. Deltas of change have the same approach.
Theory of Change:
The Equiano Institute aims to proactively shape the development and governance of artificial intelligence in Africa to ensure it benefits African nations and peoples. The key pathways are:
Research & Analysis: Conduct rigorous research to understand the impacts, risks and opportunities of AI for African contexts across areas like healthcare, agriculture, education etc.
Policy & Regulation: Develop model policies/regulations aligned with African values/priorities to govern AI development and use by nations and industry.
Capacity Building: Enhance AI governance and ethics capacities across African institutions through training, knowledge sharing and partnerships.
Public Engagement: Raise awareness and facilitate inclusive public discourse to ensure AI governance has participation from stakeholders across Africa.
Regional Coordination: Foster coordination between African nations and regional bodies like the AU to create aligned AI governance frameworks.
Applying the Delta of Change:
Identify Potential Deltas:
- Advances in AI capabilities outpacing governance frameworks
- Uneven adoption of AI across African nations/sectors
- Emergence of new AI use-cases or risks unforeseen in initial analysis
- Shifts in public opinion, political priorities or stakeholder stances
- Developments in international AI governance norms/standards
Assess Impacts of Deltas:
- How do identified deltas alter risks, opportunities or priorities?
- Do they necessitate revisions to research focus areas, policy models or capacity building efforts?
- Could they undermine public trust or political buy-in if not addressed?
Navigate Deltas:
- Continually update research agenda to analyze impacts of new AI deltas
- Be prepared to revise and evolve policy/regulation guidance based on deltas
- Proactively engage public/stakeholders on emerging deltas to maintain support
- Adapt capacity building initiatives to account for evolving delta-driven needs
- Advocate for regional governance to be iterative to deltas across Africa
Institutionalize Delta Responsiveness:
- Formalize delta-scanning and reassessment as part of institute processes
- Establish contingency plans and alternate pathways for major potential deltas
- Embed an ethos of continual learning and strategy "deformation" based on deltas
- Advocate for African AI governance itself to be a living, adaptive framework
