Alex “Sandy” Pentland is a pioneer in harnessing network science to understand and change real-world human behaviors. The director of the MIT Connection Science Research Initiative, and creator and previous director of the MIT Media Lab and the Media Lab Asia in India, Pentland recently delivered the Special Lecture at the National Academy of Engineering’s annual meeting — in which he discussed how tools like artificial intelligence and large language models could help integrate human behavior into predictive models to improve our responses to vexing societal challenges such as climate change, pandemics, and income inequality.   

Read some excerpts:  

The reason you might want to pay attention, other than sort of intellectual interest, is this: The federal infrastructure bill, the new one, includes social structure in the definition of infrastructure. In other words, they’re saying engineering social structure is now part of engineering. And that might be a little controversial, but the noncontroversial way to put it is this: When you build a highway, you advantage certain people and disadvantage other people. When you come up with any sort of policy or infrastructure, it has differential effects, some of which are not immediately obvious, and it’s saying, “You should think about that.”  

Now, why did they put that in that bill? Because of the failures that we’ve experienced from not accounting for human social structure and human behavior. 

We have perpetual problems where we don’t know what to do — inequality, climate change action, etc., etc. And many of those things hinge not on the technology or the systems that we engineer but on human behavior.  

It is a good thing to improve social exploration between communities, using knowledge of social interactions. Everything from infection rates, to innovation rates, to intergenerational mobility, all of those things depend on [social interactions] as a principal causal element. It is a complex situation of feedback loops.  

It seems to me that it is a natural way for AI to be used — to help people understand more opportunities, more facts, what other people do — as a common-sense engine to help upskill people. And it seems like that’s also something that’s a little safer than a lot of the things that people do, because it’s not taking agency from the person. It’s providing them with ... “what everybody else does is this, now, you can make up your mind about it.”  

And I think it’s something we really need to do — to have continuous learning become a standard part of our society. 

Watch the full lecture here: 

Engineering Ecosystems with AI: 2023 NAE Annual Meeting Special Lecture on Engineering and Society