Common Sense in AI

For some time, common sense has been viewed as a valuable source of knowledge in AI. Recently, a significant push is underway to accumulate a body of knowledge representative of common sense that can guide AI in a more socially-relevant manner. AI models are currently divorced from how everyday persons understand themselves, others, and their surroundings, and thus tend to hallucinate, or make inappropriate responses. The assumption is that a good dose of common sense may bring AI closer to how persons actually think and behave. Yet the search for common sense is guided by Cartesianism, and thus the resulting common sense does not reflect the daily lives of persons. In the Cartesian framework, the aim is to accumulate a large amount of objective data that allegedly represent the daily lives of persons. This essay advances an alternative, less dualistic philosophy that can bring AI closer to persons and how they make decisions. AI should be grounded in the lifeworld (Lebenswelt) of persons, a notion borrowed from phenomenology, so that this technology may become more individually and socially-attuned and relevant. FlourishingAI, a project in Bogotá, Colombia, is an example of how this philosophy can be put into practice. This essay envisions a more socially-responsible and less alienating AI that does not pose a threat to human existence.