Human-inspired mechanisms

When people draw inferences and justify their claims, they do not typically follow formal mathematical logic. Instead, people use a range of specialized problem-solving skills: analogies, stereotyping, generalization, specialization and case-based reasoning. Qualitative reasoning, Analogical reasoning, Scripts and Case-based reasoning are attempts to translate such intuitive thought processes into computational models. For example, rather than performing a precise heat-transfer computation to understand a cooling cup of coffee, a system might combine simpler models of thought to understand ‘hot’ and ‘warm’ as analogs of ‘full’ and ‘empty’ and thereby understand heat transfer as like a leaky bucket. Likewise, continuous variables may be factored along physical and psychological inflection points (solid, liquid, cold, warm, hot) to reduce search space and more closely match the intuitions of people.

Thus, these approaches benefit from a symbolic heritage, while having vastly reduced search spaces and an ability to perform meaningful (if only approximate) inferences by analogy, when a problem is not fully understood by the system.