Learning and innovation in self-governing systems
File(s)
Author(s)
Mertzani, Asimina
Type
Thesis
Abstract
The future of distributed information processing in socio-technical systems is hybrid; it involves
meaningful interactions between human (natural) and computational (artificial) intelligences. This
hybrid nature raises unique challenges for self-governance and deliberation with respect to social
arrangements (i.e. socially constructed rules, norms, structures, and procedures). To address
these challenges and enable individuals and communities to define, modify, and envision social
arrangements, we need mechanisms for learning and innovation. Accordingly, this thesis presents:
• a critical synopsis of interdisciplinary literature from psychology, economics, philosophy, and
political science, which shapes the definition of self-governing systems;
• an incremental refinement and integration of three methodologies: socially-inspired comput-
ing, hybrid reasoning and learning (neurosymbolic) systems, and socially-guided machine
learning, to produce a new methodology called Socially-Guided Reasoning and Learning for
designing and implementing self-governing systems;
• three novel algorithms for addressing challenges in self-governance:
– RTSI+ for explanatory adequacy and epistemic coherence through social influence,
– 4voices for effective self-regulation via requisite influence, and
– Θ-learning for reaching sustainable collective agreements through consensus and dissent.
By animating these algorithms with multi-agent systems, experimental results offer insight
into issues of epistemic coherence, systemic stability, and the relationship between founda-
tional goals and collective decision making;
• a proof of concept for a novel innovation-support system integrating multi-agent systems
simulation, generative AI, and human users for enabling the human-AI co-production of
social arrangements;
• a metrics-based evaluation of self-governing socio-technical systems in terms of sustainability,
performance, and human empowerment.
In conclusion, we argue that the significance of this thesis is to have produced sustainable mecha-
nisms for supporting empowerment, preserving dignity and fostering human flourishing during the
transition to a hybrid future.
meaningful interactions between human (natural) and computational (artificial) intelligences. This
hybrid nature raises unique challenges for self-governance and deliberation with respect to social
arrangements (i.e. socially constructed rules, norms, structures, and procedures). To address
these challenges and enable individuals and communities to define, modify, and envision social
arrangements, we need mechanisms for learning and innovation. Accordingly, this thesis presents:
• a critical synopsis of interdisciplinary literature from psychology, economics, philosophy, and
political science, which shapes the definition of self-governing systems;
• an incremental refinement and integration of three methodologies: socially-inspired comput-
ing, hybrid reasoning and learning (neurosymbolic) systems, and socially-guided machine
learning, to produce a new methodology called Socially-Guided Reasoning and Learning for
designing and implementing self-governing systems;
• three novel algorithms for addressing challenges in self-governance:
– RTSI+ for explanatory adequacy and epistemic coherence through social influence,
– 4voices for effective self-regulation via requisite influence, and
– Θ-learning for reaching sustainable collective agreements through consensus and dissent.
By animating these algorithms with multi-agent systems, experimental results offer insight
into issues of epistemic coherence, systemic stability, and the relationship between founda-
tional goals and collective decision making;
• a proof of concept for a novel innovation-support system integrating multi-agent systems
simulation, generative AI, and human users for enabling the human-AI co-production of
social arrangements;
• a metrics-based evaluation of self-governing socio-technical systems in terms of sustainability,
performance, and human empowerment.
In conclusion, we argue that the significance of this thesis is to have produced sustainable mecha-
nisms for supporting empowerment, preserving dignity and fostering human flourishing during the
transition to a hybrid future.
Version
Open Access
Date Issued
2024-11-01
Date Awarded
01/05/2025
License URL
Advisor
Pitt, Jeremy
Publisher Department
Department of Electrical and Electronic Engineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)