Civic AI

mediumThis post was originally published by Dark Matter at Medium [AI]

The ACI framework provides a tool for considering how the interaction between human intelligence (HI) and artificial intelligence (AI) can be applied to community-based challenges. It represents augmented collective intelligence as emerging from the interaction of people, technology, and objects in the physical world and the digital environment. People and civic assets (e.g. a smart solar panel or a tree fitted with a sensor) make up the physical context. The digital environment contains representations of the physical world, by using so-called digital twins. People and civic assets are connected to their digital twins through data streams that are continuously updated with fresh data. We refer to the digital twins of people as Personal Digital Twins (PDTs), whereas those of the civic assets are called Asset Digital Twins (ADTs).

The ACI framework allows a user to model the interactions between these components, as well as to map data and information flows. Agent-based models (ABMs) and simulations are used to explore counterfactual scenarios, taking data from the physical environment to simulate potential outcomes, which can be used to help decision-making in the real world. In Section 3 below, we will examine three examples of how the framework can be applied to community challenges.

A community’s response to the climate crisis is a hybrid system combining natural, human, and machine subsystems, each consisting of a diverse set of individual agents. Given the varied objectives and preferences of the agents, we can model the system as an ABM involving interactions between human agents, machines, and the natural environment. The ABM can be used to study the effects of individual actions by the agents on the collective outcome of the system. In this way, agent-based modelling and the concept of complex adaptive systems (CAS) can be used to understand the dynamics of wicked problems like the climate crisis. Rather than solving complex systems through the identification of optimal strategies, CAS and ABM consider the system as dynamically evolving within the constraints of the real-world environment. This modelling methodology recognises the fact that in collective action problems there may be mutually exclusive trade-offs between many stakeholders. As a consequence, the likelihood of an ‘optimal’ solution is small.

Part of the difficulty in addressing the climate crisis lies in aligning individual and collective values and belief systems. For instance, the challenge to implement the internationally agreed climate goals of the Paris Agreement through local responses consists of choosing actions that connect collective values to individually held beliefs and behaviours. We need to find better ways to align individual perspectives and collective goals. Decades of experiments on collective risk and collective benefit scenarios have exposed serious limitations in the cognitive abilities of individuals to accurately perceive risks and handle uncertainty (e.g. through probabilistic reasoning). The combination of artificial and collective intelligence can help to reconcile these tensions to incorporate more diverse collective values.

We have sought to embed consideration of AI ethics at the core of the ACI framework and three use cases. This has been influenced by a review of the issues surrounding value alignment, moral rules, and ethics in AI, and an extensive survey of ethical principles for AI and the Trustworthy AI guidelines produced by the High-Level Expert Group of the European Commission. Value alignment in AI refers to the need to make sure that AI systems act in accordance with human values. Previous attempts to crowdsource moral principles have not only demonstrated the fluid nature of moral convictions globally, but have also led to criticism of the validity of this approach. Directly encoding human values into a machine seems impossible, as illustrated by moral dilemmas such as trolley problems. After all, whose values should a given AI system be aligned to? Even within countries, different groups of people may have different value systems or the same community may set different priorities depending on contextual factors.

These choices have implications at all stages of implementing ACI, starting with the design of the user interfaces and progressing to the parameters of impact assessment algorithms. To combat this, we propose that every application of Civic AI should start with consideration of the normative aspects of value alignment. After articulating and discussing the diverse values that are surfaced, communities and institutional partners can use the framework to explore how to reflect them in the overall design. We propose an approach where community-held values determine both normative and technical aspects of the combined human-machine system.

In designing this approach, we have also drawn from value-sensitive design research on inclusivity and from community-based participatory design processes. We illustrate how some of these principles apply in practice in the next section.

The ACI framework is intended as a tool to help CSOs, local authorities, and AI experts consider augmented collective intelligence across a range of community-based applications. For now, we have focused on three use cases.

  1. Connected urban forest. Applying ACI to develop automated and participatory processes for mapping, monitoring, caring for, and measuring the holistic benefits of urban trees.
  2. Collective climate action. Adopting ACI to help communities undertake sensemaking, simulate the impact of and commit to climate positive actions.
  3. Participatory energy. Integrating ACI to help set-up, operate, maintain, and model the financial and social outcomes of community energy initiatives.

When exploring the three areas of focus, we uncovered several opportunities, where the ACI framework can help resolve the six common challenges we have identified. Civic assets form an important focal point for addressing these. Investing in shared infrastructure is critical. It will lead to the development of mechanisms that help communities deliberate, plan, implement, and measure the impact of climate crisis adaptation and mitigation projects.

  1. Collective understanding. Democratising deliberation among all community stakeholders by augmenting agency to non-humans, inanimate things, and future generations, and acknowledging their right to participate in the decision-making process.
  2. Participatory maintenance. Enabling collective caring for civic assets by augmenting their agency with sensors to identify their needs and creating new interfaces to facilitate interaction.
  3. Impact modelling. Analysing the comparative impacts of climate actions to foster informed commitments and sustained behavioural change.
  4. Commitment engines. Devising mechanisms that can help individuals and communities to adopt and sustain climate positive actions.
  5. Automated bureaucracy. Digital, replicable, and adaptable templates to replace paper-based processes using open data standards, collective and decentralised data governance, and common APIs.
  6. Agile resource balancing. Enabling distributed resource networks by increasing community agility to locally manage resource demand and supply and by more accurately forecasting future usage patterns.

Connected urban forest — Summary

Trees are crucial for our urban carbon transition. They help to mitigate the impacts of the climate crisis by providing cooling, preventing flooding, and supporting urban biodiversity, among many other benefits. Yet, cities are struggling to match the scale of planting needed to meet their net-zero targets, partly because of the difficulties of measuring these benefits, which results in them being understood as a cost rather than as an investment. ACI could help us transition to a new reality in which people and machines work collectively to map, monitor, maintain, and measure the holistic benefits that urban trees provide, thereby justifying their funding.

Cross-cutting ACI opportunities of particular relevance to this use case:

  • Participatory maintenance
  • Impact modelling
  • Automated bureaucracy
Spread the word

This post was originally published by Dark Matter at Medium [AI]

Related posts