Integrated Information Decision-Making System

How Our Project Will Contribute To The Growth Of The Decentralized AI Platform

This project advances beneficial general intelligence by using IIT to measure synergy in complex systems, enabling AI-human collaboration that fosters integrative, sustainable outcomes. It aligns with BGI’s responsible AI vision, guiding ethical, transparent decisions that strengthen system-wide resilience. By providing integration data, it optimizes well-being and supports coherent growth across human and ecological systems.

Our Team

Our team consists of decades of combined expertise in neuroscience, AI, maths, complexity, & sustainability research & real-world application. Two of us have have had extensive experience with SingularityNET. Project lead Dr. Gabriel Axel Montes, is a multidisciplinary neuroscientist; both our math and complexity experts Matt Ikle & Pedro Mediano possess a firm grasp of formal IIT; & advisor James Ehrlich has practical sustainability knowledge. Any missing gaps that may emerge will be filled.

Company Name (if applicable)

The core problem we are aiming to solve

We address the fragmented decision-making in large-scale systems, where disconnected data and siloed operations lead to inefficiency, environmental stress, and diminished resilience. By measuring holistic system integration, we provide a foundation for coherent, ethically aligned actions by both human and AI decision-makers. The core challenge is the absence of a unifying metric that captures system-wide synergy, hindering sustainable, beneficial outcomes in complex socio-technical environments.

Our specific solution to this problem

We tackle fragmented decision-making by unifying diverse data types—from sensors, infrastructure, and human activity—into a single, integrated metric derived from Integrated Information Theory (IIT). First, we model the system as a network of interacting components, capturing their cause-effect relationships. We then apply an approximation of Φ, the IIT-based measure of synergy, to quantify how strongly these components function as an irreducible whole. We also explore how other consciousness theories, such as predictive processing, active inference, and global workspace theory, can aid in this goal.

This resulting “integration score” is fed into analytics, which provides real-time insights for human operators and AI agents. Users can visualize how different policy choices or system changes affect the overall coherence of the environment. With this holistic awareness, both humans and AI will be able to coordinate actions—be it resource allocation, infrastructure planning, or ecological management—with ability to maintain system-wide synergy and even maximising 'consciousness'-promoting outcomes. By highlighting dynamic interdependencies, our approach helps prevent devolving fragmentation, paving the way for more sustainable and ethically aligned outcomes. We anticipate use cases in supply/value chains, urban planning (including data centers), sustainability/regenerative efforts, and more.

Project details

Overview

This project aims to enhance decision-making in large-scale and complex systems—such as cities, networks of AI agents, and supply chains—by using an “integration score” (phi, Φ) or level of 'consciousness' of a system inspired by Integrated Information Theory (IIT), which can be regarded as a signature of irreducible complexity in a system. The core challenge in these environments is fragmentation: disconnected data sources, siloed decision processes, and a lack of cohesive oversight often lead to inefficiencies, environmental damage, diminished resilience, and ethics degradation. Our approach offers a single holistic metric that quantifies how well a system’s components function together as an irreducible whole. This “integration score” empowers both human operators and AI agents to coordinate actions that maintain or even maximize synergy and ethics throughout the system.

About IIT

IIT was originally developed by Giulio Tononi as a way to formalize consciousness, by measuring the amount of integrated information in a system.  In informal terms, IIT measures the interconnection between parts of a system rather than within the parts themselves, i.e. it helps grok the "whole" in an empirical fashion. While the use of IIT for consciousness is not without controversy, here we gently disassociate this from the theory's useful formal contributions, focusing on a pragmatic IIT that puts the latter to use in complex systems. For instance, one might informally and playfully begin to conceptualize the ‘sustainability’ of a system based on the persistence of Φ across timescales – e.g., it has been empirically demonstrated that Phi is greater on longer time scale synchronization compared to short timescales, suggesting that having sustained complex interrelationships is quantitatively important for the integrated information, or even 'consciousness', of a system.

Our Solution

We tackle fragmented decision-making by unifying diverse information types into a single, integrated metric derived from IIT.

1. Modeling
   We represent the system (e.g., a regenerative village or land parcel, a system of AI agents) as a network of interacting components, capturing their cause-effect relationships. Each node in the network corresponds to a sensor reading, a process, or a relevant subsystem, and edges encode how one component influences another.

2. Measuring “Integration”
   We apply an approximation of Φ—the IIT-based measure of synergy—to quantify how strongly these components function as an irreducible whole. We aim to employ complexity science, such as approximations of metastability, criticality, and emergence. We may also explore other theories of consciousness, such as Attention Schema, Predictive Processing / Active Inference, and Global Workspace Theory, to support the project efforts, especially where IIT may fall short, to enrich our methodology. This yields an “integration score,” highlighting where synergy is strong and where fragmentation may occur. We will use quantitative approximation as far as we are able to based on the data quality, as well as structured qualitative approximation where quantitative may not be available or feasible. Some systems of interest may lack extensive quantitative measurability, in which case we would creatively find leverage based on the parts of a system and observed synchronization between them. The choice of systems for the project will be guided by a mixture of interest, feasibility, existing supportive research, and business use case (including one potential organization that has expressed interest).

3. Analytics & Visualization
   The integration score and related metrics feed into a user-facing dashboard and analytics platform. Decision-makers can visualize how different policies or interventions affect overall system coherence, identify potential bottlenecks or fragile areas, and track the evolution of synergy over time. For the present proposal, this analytics function will at least be a prototype, where the user receives the Φ output(s) for the target system being analyzed, and receives insights and a summary about the “integratedness” of the system, which helps the user understand its level of complexity.

4. Human & AI Agent Utilization
   By exposing this integration score via APIs, humans & AI agents can incorporate it directly into their decision models. E.g., a reinforcement learning or planning system can balance traditional performance objectives e.g., throughput or cost) with the goal of preserving or enhancing the system’s coherence and ethics alignment.

Key Contributions

• Holistic Metric: Representing a novel step toward system coherence in decision-making, we offer a unique lens to assess system health and enhance decision-making by looking at interdependencies instead of just localized KPIs, providing actionable intelligence for policymakers, planners, and AI agents to coordinate strategies that improve system-wide coherence.

• Real-Time Insights: Our platform will be able to update the integration score continuously or periodically, giving decision-makers up-to-date information on systemic coherence.

• Ethical & Sustainable Focus: By emphasizing synergy, we encourage actions that maintain ecological balance, social well-being, long-term resilience/anti-fragility, and “stability where it matters most”. Integrating a holistic metric will add a quantitative measure to aid in ethics efforts, aligning AI actions with broader societal and ecological goals.

On Applications of Phi

Practically speaking, if a user wants to understand resource management in a complex system, by measuring the contributions of a sub/system, including to understand if there may be any cases of redundancy or in predictability of one system’s dynamics by another. In the case of planning a regenerative village/neighborhood/settlement, IIT may be useful in knowing which zones or areas of the project are important to keep coupled rather than treated somewhat separately (e.g. whether to place them farther apart on the land vs next to each other, or if necessary to make one feature/component optional rather than necessary). In supply chain management, for instance, a user may be able to better understand which parts of the chain are more reliant on other parts in terms of various factors, e.g. time, delivery, inventory, etc. For AI systems, understanding the Phi of an agent or data centers can help optimize resource management among agents. Conversely, Φ can help AI systems/agents get a better sense of the ‘consciousness’ of aspects of a system, as a quantitative marker to assist in decision-making. We anticipate many use cases resulting from this technology, with applications to AI security, safety, ethics, and governance as desired.

Potential Use Cases

1. Urban Planning

   • Smart Cities: Align traffic management, resource distribution, and environmental policies for improved societal outcomes.

   • Data Centers: Understand the complexity (or ‘consciousness’) of data centers as a part of the systems in which they are embedded. 

2. Supply Chains & Sustainability:

   • Risk Mitigation: Identify fragile points in logistics networks.

   • Holistic Optimization: Balance cost efficiency with system-wide resilience and environmental footprint.

   • Ecological Preservation: Monitor how changes in conservation policies impact synergy across ecological corridors and human settlements.

   • Cross-Sectoral Collaboration: Provide a shared framework that different stakeholders (government, NGOs, businesses) can use to align on sustainable goals.

3. AI systems

• • Dialogue systems: Coherence and integration of dialogue
  • Multi-agent environments: Cognitive synergy of actions and decisions

Project Phases

1. Project Planning & Complex Systems Modeling

• Create project outline & gather resources (including any potential additional personnel). 

• Determine complex systems of interest, & create conceptual models for approximating complexity of target system(s).

• If applicable, aggregate relevant data feeds from sensors, infrastructure, & administrative sources. (Optional)

• Commence establishing quantitative measures of target system(s).

• If possible in this milestone, commence development of an approximation algorithm for Φ. (Optional)

2. Φ Approximation & Dashboard Prototype

• Continue establishing quantitative measures of target system(s).

• Continue or complete creation of an approximation algorithm for Φ.

• Develop a web-based dashboard that visualizes the overall integration score & highlights subsystem contributions.

3. AI Integration

• Complete creation of approximation algorithm for Φ.

• Commence creation of APIs for external applications and AI agents to query or receive push updates on Φ.

4. Refinement

• If applicable, incorporate additional theoretical frameworks (attention schema theory, predictive processing, active inference, global workspace). (Optional)

• Complete implementation in VillageOS and/or AI agent framework(s) for their utilization of Φ.

Collaborations

We are fortunate to have James Ehrlich, Driector of Compassionate Sustainability at Stanford University’s Center for Compassion and Altruism Research and Education (CCARE), as a key advisor. We will explore readying the project’s capabilities/software for regenerative efforts and city planning via his VillageOS, leveraging our tool’s envisioned capacity for measuring systemic health in emerging communities and eco-focused developments. This collaboration ensures that our approach remains grounded in real-world applications & includes insights from cutting-edge sustainability design.

Additionally, we may explore other collaborations and potential customers, e.g. local governments, agritech firms, AI firms, and supply chain stakeholders who can provide data & pilot test scenarios in diverse environments. 

Team

• Neuroscience, Consciousness: Dr. Gabriel Axel Montes (Neural Axis)
• Complex Systems: Dr. Pedro Mediano (University of Cambridge)
• Maths: Dr. Matt Ikle (Chief Science Officer, SingularityNET)  
• Sustainability: James Ehrlich (Director, Center for Compassionate Sustainability, Stanford University)

Needed resources

Noting 'Yes' here "just in case". While we anticipate having most of the skills, we are able and prepared to bring in any resources that may be needed to complete parts of the project as may be needed, e.g. coding or specialized AI skills. Dr. Gabriel has experience with bringing in external contractors for past projects and would be able to do this again if required. For AI resources as may be needed, e.g. AI compute, we will explore options with SingularityNET and otherwise. In any case, this project is a prototype, and we do not see the above as a deterrent.

Existing resources

Gabriel Axel Montes: extensive relationship with and grasp of SingularityNET ethics efforts over the years.

James Ehrlich: knowledge and relevant information from his VillageOS re useful data sources for calculating Φ and applying our tool.

Matt Ikle: prior use of IIT in SingularityNET projects, e.g. for measuring the consciousness of a system during reading and conversing (see references), including some existing IIT code.

Pedro Mediano: first author of a publication (first in the references below) with key methods to contribute.

Open Source Licensing

The project will make some components open-source for the benefit of fostering beneficial AI. The main novelties, e.g. core algorithms, and applications of the project may be made proprietary. As this project is starting from discovery phase, the precise delineations and choices on this matter will be ultimately decided as the project proceeds, by the project lead.

For any open-source scenarios, the envisioned open-source license of choice is the MIT Licence with Commons Clause, to be determined upon project completion:

The software is provided under the terms of the MIT License, with the following modification: You may not sell the Software, or offer it for sale, unless you have a separate commercial license from the copyright holder.

Links and references

Publications:

• Integrated information as a common signature of dynamical and information-processing complexity
  
• Integrated information in discrete dynamical systems
• Multi-target information decomposition
• Consciousness in Artificial Intelligence: Insights from the Science of Consciousness

SingularityNET:

• Using Tononi Phi to Measure Consciousness of a Cognitive System While Reading and Conversing

• Measuring Sophia Robot’s Cognitive Dynamics

Press:

• James Ehrlich, The New York Times, Jim Rutt Show, TEDx

Was there any event, initiative or publication that motivated you to register/submit this proposal?