Research Interests

The Emergence of Social Learning in Artificial Evolution Systems

When studying social animals we are essentially taking a top down view on social learning; we are can understand how social learning manifests itself but we can't really tackle the question of how the social behaviors we are observing emerged as viable adaptive strategies. Whilst artificial evolutionary systems, like all models, are limited, they do enable us to asses not only which social behaviors evolve, but also how and when they evolve.

I am interested in tackling the question of How can social learning evolve from the bottom up. To investigate this problem I create grounded, neuroevolutionary, artificial life models in which populations of artificial agents evolve over time to tackle simple foraging problems. The agents have access to only simple social/public information, such as the age or health of the other agents. From this simple social information a variety of complex social learning strategies evolve in response to the difficulty of the environmental and behavior of other agents. Going forward I intend to make use Neuromodulated Plasticity and NEAT to allow for complex learning and the evolution of neural networks topologies.

Related Publications:

• Borg JM and Channon A. (2021). The effect of social information use without learning on the evolution of social behavior. Artificial Life, vol. 26(4), pp.431-454. MIT Press. https://doi.org/10.1162/artl_a_00328

• Borg JM and Channon A. (2017). Evolutionary Adaptation to Social Information Use Without Learning. Applications of Evolutionary Computation, EVOAPPLICATIONS 2017, PT I (vol. 10199, pp. 837-852). https://doi.org/10.1007/978-3-319-55849-3_54

• Borg JM, Channon A, Day C. (2011). Discovering and maintaining behaviours inaccessible to incremental genetic evolution through transcription errors and cultural transmission. In: ECAL 2011: Proceeding of the Eleventh European Conference on the Synthesis and Simulation of Living Systems. (pp. 101-108). MIT Press. https://doi.org/10.7551/978-0-262-29714-1-ch019

The Development of Social Learning Strategies in Populations of Evolving Artificial Agents

In nature we see a variety of social learning strategies being employed, from copies the most common behavior (conformist social learning), to unbiased copying and even anti-conformist learning. I am interested in how these strategies can develop in response to environmental variability and environmental difficulty. By using populations of evolving artificial agents I hope to investigate when different social learning evolve, and when hybrid strategies evolve in populations. I also hope to see how important non-social learning (individual learning/asocial leaning/innovation) is important for enabling the evolution of different strategies - do some social learning strategies require more of less individual learning than others do be viable in the long term?

Related Publications:

• Jolley BP, Borg JM, Channon A. (2016). Analysis of Social Learning Strategies When Discovering and Maintaining Behaviours Inaccessible to Incremental Genetic Evolution. From Animals to Animats 14 (vol. 9825, pp. 293-304). http://dx.doi.org/10.1007/978-3-319-43488-9_26

• Borg JM, Jolley BP, Channon A. (2016). Social Learning Strategies: Who you learn from affects how new behaviours are discovered. The First International Workshop of Social Learning and Cultural Evolution. Cancun, Mexico. SLaCE-2016

Robust Social Behaviors in Evolving Robots

Over the next few decades robots (and other artificial agents) are expected to become an increasingly common part of modern life. However, if these robots are to become robust and autonomous parts of our day to day life they are going to need to learn and interact socially, both with Humans and one another. As it currently stands social behavior is currently built in to social robots, with very little on-line and social learning taking place - these behaviors are not only brittle in the long term, requiring constant updating, but are also biased by their human developers. One approach to making social robots socially robust and reactive to allow robot social behaviors to evolve over time with minimal developer input - this could lead to emergent robot social behavior and even robot culture. I am interested in the questions of how can be evolve robot social behavior that is still functional, safe and ethical. These questions are from both a scientific perspective: "how do we actually do evolve robot social behavior?", and a philosophical perspective: "what would acceptable robot social behavior look like?"

Open-Ended (Cumulative) Cultural Evolution

Details forthcoming

Potts' Variability Selection Hypothesis: Social Learning and Culture as a response to temporal environmental change

Paleoanthropologist Rick Potts has posited that Human evolution, including our ability to adapt rapidly to changing environments, our culture, and our social systems, are a direct result of what may be described Variability selection. According to Potts Variability selection "is a process considered to link adaptive change to large degrees of environment variability. Its application to hominid evolution is based, in part, on the pronounced rise in environmental remodeling that took place over the past several million years." (https://bit.ly/2PN6GzU)

I am interested in testing Potts' Variability Selection Hypothesis in artificial evolution systems, and investigating the link between temporal variability and the evolution of social behaviour and emergence of culture. The core question I am interested in investigating is what are the necessary temporal environmental conditions to give rise to the evolution of advanced social behaviours leading the emergence of culture?

Related Publications:

• Borg J and Channon AD. (2012). Testing the Variability Selection Hypothesis: The Adoption of Social Learning in Increasingly Variable Environments. In: ALIFE 13: The 13th Conference on the Synthesis and Simulation of Living Systems. (pp. 317-314). MIT Press. https://doi.org/10.7551/978-0-262-31050-5-ch042

Coloured Noise and the Evolution of Environmental Tolerance

Details forthcoming

Related Publications:

• Grove M, Borg JM and Polack F. (2020). Coloured noise time series as appropriate models for environmental variation in artificial evolutionary systems. Proceedings of ALIFE 2020: The 2020 Conference on Artificial Life. (pp. 292-299) MIT Press. https://doi.org/10.1162/isal_a_00284

(The Evolution of) Sociotechnical Systems

I am interested in how Humans interact socially with systems (such as smart energy meters, autonomous cars, robots), and how these systems interact with each other in a social manner. Ultimately I am interested in seeing if new social behaviours can emerge within and between sociotechnical systems.

Related Publications:

• Brooks N, Powers ST and Borg JM. (2020). A mechanism to promote social behaviour in household load balancing. Proceedings of ALIFE 2020: The 2020 Conference on Artificial Life. (pp. 95-103) MIT Press. https://doi.org/10.1162/isal_a_00290

Evolutionary Art

Can we evolve artwork without the need for an human artist, and can artificial intelligence every be truly creative? I am interested in evolving artwork in a way that keeps on producing new and interesting artwork and new interesting styles - this could be described as open ended evolutionary art. I am yet to pursue this interest academically, but would be very interested in being contacted by potential PhD students who are interested in exploring open ended evolution through the medium of evolutionary art.

Neuroevolutionary Approaches to Predicting Election Results

Predicting the results of elections is extremely challenging; polling companies, media outlets, and political scientists rarely predict the outcomes of elections, with each new election giving rise to unpredicted and maybe unpredictable outcomes. I am interested in exploring a neuroevolutionary approach to the problem whereby neural networks are evolved to predict the outcomes of elections based on census data, polling data, and previous election results.

Electoral Signatures

Details forthcoming

Fair Digital Societies

Details forthcoming