Indicator Selection and Weighting as a Human-GAN Model
[Continuation of the discussion from chapter 8]
The modelling of social and global entropy, as proposed by the author, offers a promising approach for quantifying complex sustainability challenges. Nevertheless, the model is not without significant limitations. This paper examines how the Human-GAN model devised by Fon can assist in overcoming these limitations, thereby establishing a more robust foundation for entropy-based sustainability research.
Limitations of the original entropy model
Fon’s original model for quantifying social and global entropy has several weaknesses:
a) Subjectivity in indicator selection and weighting
b) Loss of detailed information through aggregation
c) Challenges in data collection and quality
d) Questionable transferability of the physical entropy concept
The Human-GAN model as a solution approach
The author proposes a Human-GAN (Generative Adversarial Network) model that enables co-evolution between humans and AI to refine the entropy concept. This model includes:
• An expert commission as a ‘generator’
• A swarm rating agency as a ‘discriminator’
• Citizens’ councils as curators
Overcoming subjectivity through participatory processes
The Human-GAN model addresses subjectivity in the selection and weighting of indicators:
• Iterative feedback loops between experts and the public
• Integration of different perspectives through citizens’ councils
• Dynamic adjustment of indicators based on swarm feedback
This approach is in line with concepts of participatory modelling (Voinov et al., 2016) and can lead to a broader acceptance and validity of the model.
Preservation of detailed information through multi-level aggregation
In order to offset the reduction in detailed information, the Human-GAN model could be employed.
• Introduce multiple aggregation levels
• Develop context-specific sub-indices
• Provide interactive visualisations for exploring different levels of detail
These approaches facilitate the identification of both overarching trends and specific aspects of social entropy (Liu et al., 2019).
Improving data collection through citizen science
The integration of citizen science approaches into the Human-GAN model could improve data collection and quality by:
• Crowdsourcing of local data
• Validation of official data by citizens
• Development of citizen sensor networks
These methodologies have been demonstrated to be effective in other domains of environmental research (Bonney et al., 2014), and they have the potential to considerably augment the database for the entropy model.
Further development of the entropy concept
In order to improve the transferability of the entropy concept, it is proposed here:
• Integration of information-theoretical entropy concepts (Shannon, 1948)
• Development of hybrid entropy models that combine physical and informational aspects
• Utilisation of the Human-GAN for iterative refinement of the entropy concept
This further development may facilitate a more comprehensive understanding of social and global entropy that considers both physical resource flows and the informational aspects of social systems.
References:
Bonney, R., Shirk, J. L., Phillips, T. B., Wiggins, A., Ballard, H. L., Miller-Rushing, A. J., & Parrish, J. K. (2014). Next steps for citizen science. Science, 343(6178), 1436-1437.
Fon, U. (2024). Künstliche Superintelligenzen und göttliche Vorstellungswelten. In “Von der Kunst der Imagination zur Vision künstlicher Superintelligenz”. [Unveröffentlichtes Manuskript].
Liu, H. Y., Kobernus, M., Broday, D., & Bartonova, A. (2014). A conceptual approach to a citizens’ observatory–supporting community-based environmental governance. Environmental Health, 13(1), 107.
Shannon, C. E. (1948). A mathematical theory of communication. The Bell system technical journal, 27(3), 379-423.
Voinov, A., Kolagani, N., McCall, M. K., Glynn, P. D., Kragt, M. E., Ostermann, F. O., … & Ramu, P. (2016). Modelling with stakeholders–next generation. Environmental Modelling & Software, 77, 196-220.