Researchers at MIT and the University of Washington have developed a new method for modeling the decision-making behaviors of agents, taking into account computational constraints. This model, which can predict future actions from past behavior, aims to improve AI systems’ collaboration with humans by understanding and adapting to human irrationalities and decision-making processes. Credit: SciTechDaily.com
A new technique can be used to predict the actions of human or AI agents who behave suboptimally while working toward unknown goals.
MIT and other researchers developed a framework that models irrational or suboptimal behavior of a human or AI agent, based on their computational constraints. Their technique can help predict an agent’s future actions, for instance, in chess matches.
To build AI systems that can collaborate effectively with humans, it helps to have a good model of human behavior to start with. But humans tend to behave suboptimally when making decisions.
This irrationality, which is especially difficult to model, often boils down to computational constraints. A human can’t spend decades thinking about the ideal solution to a single problem.
Development of a New Modeling Approach
Researchers at MIT and the University of Washington developed a way to model the behavior of an agent, whether human or machine, that accounts for the unknown computational constraints that may hamper the agent’s problem-solving abilities.
Their model can automatically infer an agent’s computational constraints by seeing just a few traces of their previous actions. The result, an agent’s so-called “inference budget,” can be used to predict that agent’s future behavior.
Practical Applications and Model Validation
In a new paper, the researchers demonstrate how their method can be used to infer someone’s navigation goals from prior routes and to predict players’ subsequent moves in chess matches. Their technique matches or outperforms another popular method for modeling this type of decision-making.
Ultimately, this work could help scientists teach AI systems how humans behave, which could enable these systems to respond better to their human collaborators. Being able to understand a human’s behavior, and then to infer their goals from that behavior, could make an AI assistant much more useful, says Athul Paul Jacob, an electrical engineering and computer science (EECS) graduate student and lead author of a paper on this technique.
“If we know that a human is about to make a mistake, having seen how they have behaved before, the AI agent could step in and offer a better way to do it. Or the agent could adapt to the weaknesses that its human collaborators have. Being able to model human behavior is an important step toward building an AI agent that can actually help that human,” he says.
Jacob wrote the paper with Abhishek Gupta, assistant professor at the University of Washington, and senior author Jacob Andreas, associate professor in EECS and a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL). The research will be presented at the International Conference on Learning Representations.
Modeling Behavior
Researchers have been building computational models of human behavior for decades. Many prior approaches try to account for suboptimal decision-making by adding noise to the model. Instead of the agent always choosing the correct option, the model might have that agent make the correct choice 95 percent of the time.
However, these methods can fail to capture the fact that humans do not always behave suboptimally in the same way.
Others at MIT have also studied more effective ways to plan and infer goals in the face of suboptimal decision-making.
To build their model, Jacob and his collaborators drew inspiration from prior studies of chess players. They noticed that players took less time to think before acting when making simple moves and that stronger players tended to spend more time planning than weaker ones in challenging matches.
“At the end of the day, we saw that the depth of the planning, or how long someone thinks about the problem, is a really good proxy of how humans behave,” Jacob says.
They built a framework that could infer an agent’s depth of planning from prior actions and use that information to model the agent’s decision-making process.
The first step in their method involves running an algorithm for a set amount of time to solve the problem being studied. For instance, if they are studying a chess match, they might let the chess-playing algorithm run for a certain number of steps. At the end, the researchers can see the decisions the algorithm made at each step.
Their model compares these decisions to the behaviors of an agent solving the same problem. It will align the agent’s decisions with the algorithm’s decisions and identify the step where the agent stopped planning.
From this, the model can determine the agent’s inference budget, or how long that agent will plan for this problem. It can use the inference budget to predict how that agent would react when solving a similar problem.
An Interpretable Solution
This method can be very efficient because the researchers can access the full set of decisions made by the problem-solving algorithm without doing any extra work. This framework could also be applied to any problem that can be solved with a particular class of algorithms.
“For me, the most striking thing was the fact that this inference budget is very interpretable. It is saying tougher problems require more planning or being a strong player means planning for longer. When we first set out to do this, we didn’t think that our algorithm would be able to pick up on those behaviors naturally,” Jacob says.
The researchers tested their approach in three different modeling tasks: inferring navigation goals from previous routes, guessing someone’s communicative intent from their verbal cues, and predicting subsequent moves in human-human chess matches.
Their method either matched or outperformed a popular alternative in each experiment. Moreover, the researchers saw that their model of human behavior matched up well with measures of player skill (in chess matches) and task difficulty.
Moving forward, the researchers want to use this approach to model the planning process in other domains, such as reinforcement learning (a trial-and-error method commonly used in robotics). In the long run, they intend to keep building on this work toward the larger goal of developing more effective AI collaborators.
Reference: “Modeling Boundedly Rational Agents With Latent Inference Budgets” by Athul Paul Jacob, Abhishek Gupta and Jacob Andreas, ICLR 2024.
OpenReview
This work was supported, in part, by the MIT Schwarzman College of Computing Artificial Intelligence for Augmentation and Productivity program and the National Science Foundation.
Quick – roll this out and tell us what Netanyahu will do next!
Hopefully he’ll have a stroke or something.
…and I’m sure nothing will ever go wrong with this tech!
What a bloody awful idea. Time this sort of research was abolished. As for Netanyahu; the genocide will continue. One doesn’t need artificial intelligence to see that! I am reminded of a certain German’s behaviour, beginning in 1933.
“Suboptimal”? WTAF! 🤨
Suboptimal? Hamas’ killing 1,200 civilians and taking 200+ more hostage (plus torturing and raping them) is a shining example of suboptimal. They knew darn well Israel was not going to ignore them. The blame belongs with Hamas, not the Israeli response to them. By population ratio, this would be the same as killing 35,000 American civilians, and you can bet your last dirhan/dollar/renmibi/whatever that any country that did this to the US would be a radioactive parking lot in about ten minutes. Blaming Israel is blaming the victim. Don’t fall for propaganda that claims otherwise.
One last point…. Americans kill Americans every day in this country we killl children children kill children. We murder each other at a rate 25,000 a year . but that’s OK. Apparently as long as we are killing each other, we don’t have a problem with it because we need our guns no matter what. America became a superpower by murdering @ 200,000. So yes, we are way capable of exacting punishment way over the top for our own revenge and let’s call it what it is.revenge
so yippee ! yay America ! Out claim to fame is dropping a bomb and then on the other side of the mouth telling the rest of the world you can’t have the same thing inferring we’re the only ones that are allowed to kill and decide who gets to be killed.
we have no place to judge anyone else and we should stay out of it. Mind our own business put the money towards the educating of our youth so we can have proper debates.
Just to rule out the typical go to labels everyone throws at each other assuming you are either anti-this or anti-that blah blah blah I will State I come from a Jewish grandfather and Lebanese grandmother I’m an American citizen. I’m a French citizen and I have a Native American daughter so, this is not come from any bias and I have no skin in this game believe it or not. There are some people out here that can think critically keep open minds and process information and listen to other people speak if they make sense and they come from an educated, intelligent Non-biased judgmental and un hypocritical place
That is a long disclaimer, but it needs to be said because people like me can actually speak the truth and see it from both sides and for what it is like every human, I have flaws but admitting mistakes and being truthful and seeing the truth and calling it out is not one of them Any response to this that comes back as subjective or dismissive or upset is because someone is feeling offended and those cases you just cannot ration with anybody which is most of the problem today in the world anyway I digress
The fact of the matter is if America is a friend of Israel and Israel is a friend to America how do friends behave if your friend had two members of their family taken one killed and one held hostage and the response was to kill an entire town of people Because the kidnapper and murderer was in that community that is called revenge that is not called Justice a life is worth a life and every time Isreael makes excuses by saying humus is hiding amongst its people so there’s nothing they can do about it and Thomas‘s fault that is hypocrisy and manipulation if Isreael was a friend to America, they would not drag us into this and they would allow different How is America going to continue as a country and a democracy when we look for reasons to hate each other I’m absolutely disgusted and yes, that is a personal opinion, but it is with the hypocrisy of humans and the hypocrisy of leaders no Isreael maybe sick and tired of being scared to exist they may have long term trauma and sure. Harmless is made up of terrorists however, if Isreael actually did care about their people, how is this tactic helping you did not negotiate and every time you say you have hostages missing and you starve other children it appears and I’m just saying it appears that you believe your lives are worth more than others And when you bully your friend into doing what you need done your dirty work and to rope them in to something that can’t be avoided, and that Friend happens to be a pushover to peer pressure every time the wind blows like the US it appears to me that we are that rich, awkward friend that comes to school with extra lunch money And people want to be our friends so they can have what we brought and the kid wants to be popular so they let people bully them. There’s no shortage of narcissistic leaders in this world that will use any excuse to destroy humans but if Isreael wants sympathy every time they show us backstory on it hostage And expect normal people who are not involved in this to then feel absolutely nothing when we see tortured dying, starving and chilled children. Well that is neither a friend that is not someone that cares about any people and certainly not someone who is thinking about the consequences that is saying, I will scorch the Earth and take everyone down with me , and by the way, I will not do the same for my friend and I will not listen to my friend, and I will start badmouthing them and smearing their name. If you will go against me and give money to my friend Ukraine, eventually wherever that money is coming from, will run out , somewhat intelligent human and one that is not skilled in military or politics to the degree, that would apparently satisfy the vetting to be a legal expert on this. It’s pretty darn, still obvious that it’s a matter of time till people get involved who the hell are we or anybody to tell any country what to do or tell any friend what to do or to bully them , two things can exist at the same time as a human being cannot want children killed and pissed off to be labeled anti-this anti-that because of it that is gaslighting that it’s most destructive form what is painfully obvious is that nobody wins except perhaps Russia and China because if I were those leaders sitting here watching us fight about something else yet again I would shake my head Close the book and rest assure that’s only a matter of time before we destroy ourselves from the inside out people want what they want. They don’t have to destroy it and guess who gets destroyed Jewish Muslim Christian black who cares everybody humans get destroyed colleges were supposed to be a safe place and a place where ideas were shared , I do not see America democracy people are not allowed to protest peacefully everyone gaslight each other all quite pathetic. We have not figured out yet how to live amongst each other and cut the crap. Then we all deserve whatever nuclear bomb s*** hit all of us so we should just get ourselves up stop whining or get off the goddamn globe