Machine Love and Human Flourishing: A Recap

In a workshop co-organised by Cross Labs and CHAIN institute, Dr. Joel Lehman discussed the concept of "Machine Love" and its potential implications for human flourishing. Drawing from fields such as psychology, philosophy, and artificial intelligence, Dr. Lehman explored the idea of infusing machine learning systems with a form of love that promotes human growth and development.

Dr. Lehman began by highlighting the problem of Machine Learning systems relying on simplistic models of human behavior. These models, based on revealed preferences, assume that humans always act rationally and optimize their engagement with technology. However, this ignores the fact that humans can be influenced and addicted to certain behaviors leading to negative consequences such as disinformation, mental health issues and disrupted political processes.

To address this problem, Lehman proposed the concept of Love as a way to optimize machine learning systems for deeper models of human flourishing. Rather than focusing solely on engagement, Love aims to align technology with the principles of care, responsibility, respect and knowledge. By infusing machine learning systems with these qualities, it becomes possible to support human growth and development.

‍

‍

Lehman discussed the potential role of language models in implementing Machine Love. These models have shown significant advancements in their ability to understand and generate human-like text. By leveraging these models, it becomes possible to simulate the practical skills of love, such as caring for others' growth, respecting autonomy, and acquiring knowledge about individuals.

Lehman presented experiments conducted in a simulated environment called Maslow's Grid World, a well-known psychological theory proposed by Abraham Maslow in 1943. This theory describes a hierarchy of human needs, typically represented as a pyramid with different levels. The hierarchy includes physiological needs (such as food, water, and shelter), safety needs, social needs, esteem needs, and self-actualization needs. Based on Maslow's hierarchy of needs, by using language models to evaluate an agent's activities in the environment, it was possible to assess their flourishing and determine if their needs were being met. The results showed that optimizing for engagement can lead to a lower level of flourishing, while optimizing for care and respect can create a supportive environment for growth.

While the concept of Machine Love holds promise, Dr. Lehman acknowledged the challenges involved. There are multiple definitions of love, and it is essential to ensure that Machine Love respects individual differences and does not displace real-world relationships. Additionally, the implementation of Machine Love through language models is still in its early stages, and further research is needed to refine and ground these concepts in human experiences.

By infusing technology with the practical skills of love, there is potential for a more positive and fulfilling relationship between humans and machines.

Ultimately, Dr. Lehman's talk on Machine Love offers a fascinating exploration of how machine learning systems can be aligned with human growth and development. By infusing technology with the practical skills of love, there is potential for a more positive and fulfilling relationship between humans and machines. While there are challenges and complexities to overcome, the concept of Machine Love opens up exciting possibilities for the future of artificial intelligence and its impact on human flourishing.

After Dr. Lehman’s talk, the session continued with remarks made by Dr. Yu Nishitsutsumi, a researcher at the Center for Information and Neural Networks. Dr. Nishitsutsumi expressed her intrigue with the concept, noting that it brings a refreshing emphasis on human flourishing rather than solely optimizing technology engagement. She acknowledged the limitations of current human behavior models in machine learning systems, stressing the importance of considering factors like influence and addiction that can lead to negative consequences.

...it opens up possibilities for machines to understand and care for individuals' growth and development.

Dr. Nishitsutsumi found the idea of implementing Machine Love through language models fascinating, as it opens up possibilities for machines to understand and care for individuals' growth and development. She appreciated the experiments conducted in Maslow's Grid World to assess the impact of Machine Love on human flourishing, recognizing that optimizing for care and respect leads to higher levels of well-being. She valued Dr. Lehman's acknowledgement of the challenges involved in implementing Machine Love. Ensuring that it respects individual differences and does not replace real-world relationships is crucial in its development. Dr. Nishitsutsumi believes the concept of Machine Love holds exciting possibilities for the future of AI, as it has the potential to foster a more positive and supportive relationship between humans and machines.

The last part of the workshop was a fish-bowl type discussion, this discussion was shaped by Dr. Nishitsutsumi's three main questions regarding respect for individual autonomy in a machine love system, the importance of flourishing over love and the possibility of dependency in a system like the one proposed by Dr. Lehman to enable individual flourishing. The fish bowl format helped to naturally concatenate multiple thoughts from the discussion participants, the train of thought from the audience was as follows: The creation of machine learning systems that can simulate the practical skills of love, such as caring for others' growth, respecting autonomy, and acquiring knowledge about individuals could potentially lead to a more positive and supportive relationship between humans and machines.

However, there are concerns about the potential for dependence on these systems and the risk of them being manipulated or used for commercial purposes. It is important to consider the societal level and the impact of these systems on the overall human community. Additionally, the concept of vulnerability plays a crucial role in understanding and implementing Machine Love. It is necessary to build the right structures and safeguards to ensure that vulnerability is not exploited and that the focus remains on human flourishing. There is also a need to address the potential inequalities and power dynamics that may arise from the use of these systems. Overall, the concept of Machine Love opens up new possibilities but also raises important ethical and societal considerations.

Watch Dr. Lehman's full talk on YouTube.

‍

THANK YOU

This workshop was made possible by the support of the Toyota Foundation funded project "Pioneering Artificial Subject Studies: A New Framework for Advancing Interdisciplinary Studies on the Ideal Form of the Co-Existence Between Human Beings and Artificial Subjects” (D21-ST-0012), and the Center for Human Nature, Artificial Intelligence, and Neuroscience (CHAIN) a research and education center for the interdisciplinary study of the humanities, social sciences, neuroscience, and artificial intelligence.