AAAS Fellow Eugene Santos, Jr. is so interested in conversations between people that he wants to teach machines to appreciate and analyze them just like humans do.
Santos’ research is a data-driven approach to the actions that we make and the results that stem from those actions that we take. He collects data, enters it into a computer and creates models of the observed intentions of human beings. In essence, Santos’ models take observable conversations and interactions with people and attempt to predict what will come next based on what has already been observed. In this sense, the models take the narrative of what already happened and attempt to write an ending for it.
The quest began when Santos received his doctorate in computer science from Brown University. He is currently Professor of Engineering at the Thayer School of Engineering and Adjunct Professor of Computer Science at Dartmouth College, where his work on artificial intelligence intersects the areas of information, cognition, human factors and mathematics. From an early age, Santos’ parents, who had electrical engineering and computer science backgrounds, instilled these skills and interests into him.
Later on, his interest in technology and AI led him to study air force operations planning as well as search engine information retrieval.
“At the time [around 1998] we were seeing the growth of search engines, I wanted to come up with better recommender systems to recommend better websites–that’s when I was working with intentional user modeling,” Santos says.
Using computers to model human mental processing, researchers like Santos can predict the goals and intentions of a user in order to understand and ultimately provide proactive assistance with user tasks like information gathering. With the help of machine learning, this work has been applied to intelligent information retrieval and enhancing the effectiveness of intelligence analysts such as those employed by the U.S. government. This can narrow down the results of a search engine or provide greater focus to user queries. In his work, Santos examines a wide range of factors that influence a person’s behavior and uses the theory of probability to assess, quantify, and rank different degrees of influence on that person.
“You can’t say a cultural experience will always produce a particular outcome,” says Santos, “But influence can make an outcome more or less likely, so I try to capture those elements of what’s more likely and what’s less likely. That gives me a baseline. Then once I see an action, I can go back through the influence structure, including what they’ve told me about their beliefs, their demographics, their personal history, to see how they got from their background to their final action.”
Santos’ current focus is on computational intent, dynamic human behavior, and decision-making with an emphasis on learning nonlinear and emergent behaviors and explainable AI. A conversation between any two people, for example, is essentially them trying to figure out each other’s intentions. A person has the option to respond truthfully, or the conversation can include misinformation, miscommunications or unknown intentions from one or both respondents.
“I’m trying to figure out your intentions so I can better listen or better respond to whatever you’re doing,” Santos notes. “That leads to a whole set of different options such as deception, such as misinformation, etc. So almost looking at what drives us, is there a reward structure behind what we do.”
At the end of the day, computers are not quite capable of the critical thinking or the ability to connect the dots that a human is capable of for instance. It is for this reason that Santos works on collaborative teams including psychologists as well as engineers. He also lends his expertise in deciphering user intent within the STEM community. As a member of the newly established AAAS Community Advisory Board, Santos helps foster meaningful conversations and relationships on the online platform.
Whether he’s volunteering or conducting research, Santos is constantly aware of how psychology mixes with technology, and how people’s inherent biases and their intentions also play a role in scientific research.
“In any field, you can cherry-pick pieces of information which do not really showcase the entirety of the results,” he says. “You can take them out of context and you can misuse them. For researchers, you need to see the full picture. What are they talking about? And again, that’s the measure of intentions. When you completely take something out of context, is there something you learn from there? Why is it taken out of context? And what does it mean that it was taken out of context? That opens up new questions for us to explore.”