MIT: AI Can Learn Human Reasoning

New MIT research indicates that AI may become more flexible in its thinking when exposed to human reasoning. A report from the university’s Sloan School of Management covers MIT’s studies of agentic AI, including how these digital entities can be trained to reason and collaborate more like humans.

In one study, people and AI were given the same scenario: Purchase flour for a friend’s birthday cake with $10 or less, but the flour costs $10.01. Ninety-two percent of people bought the flour, but AI models chose not to, deciding the price was too high.

Ju said that with the status quo, models do exactly what they are told. However, the technology is increasingly being used in situations where that is not the right thing, and exceptions come into play. Paying an extra penny makes sense for a cake, but an extra cent per item would not make sense for a company like Walmart purchasing a large number of items.

The research found that AI’s strict adherence to rules could be relaxed when exposing models to human reasoning, allowing them to be more flexible in scenarios like hiring and customer service. While companies have embraced GenAI, widespread use of agentic AI is not yet a reality.

PYMNTS wrote that human intervention remains a core component of most AI applications across the goods, technology and services industries. While GenAI can support ideation and offer data-driven suggestions, it does not produce breakthrough innovations independently that COOs feel comfortable putting into motion.

Functions such as generating feedback on product processes, cybersecurity management and product innovation still rely heavily on human guidance, especially for technology companies. Most GenAI tools remain tethered to humans because most enterprise functions are complex, interdependent and context-rich.