A specialized chatbot named Noora is helping individuals with autism spectrum disorder practice their social skills on demand.
Knowing what to say when co-workers tell you about their weekend is a social skill that many take for granted, but for some individuals with autism spectrum disorder (ASD), this social interaction can be challenging. Struggling to find the correct response in social situations like these can negatively impact social and professional relationships for people with ASD, and can worsen co-occurring conditions like depression.
Research has shown that practicing social interactions with professionals in a clinical face-to-face intervention can improve outcomes for individuals, but these solutions are often costly or not widely available. Lynn Koegel, a clinical professor of psychiatry and behavioral sciences at Stanford University, in collaboration with Professor Monica Lam from Stanford’s Computer Science Department, are the authors of recent research published by the Journal of Autism and Developmental Disorders that investigates the role of AI in filling this gap.
“Our research has shown that face-to-face work does help with social conversation … so we wanted to see if we could translate that into computer use,” said Koegel. “Accessibility is really important because a lot of people don’t have access to a face-to-face provider and the providers can be really expensive.”
Introducing Noora
In this work, funded in part by a seed grant from the Stanford Institute for Human-Centered AI and the Kind World Foundation, Koegel and colleagues evaluated interactions between participants with ASD who struggle with social interactions and Noora, a chatbot built with a large language model (LLM).
In one-on-one interactions, which can be written or spoken, Noora offers individualized guidance on a number of social communication scenarios; it helps users learn to ask questions, give compliments, respond empathically, and with other areas of social communication that are often challenging.
In this recent work, Koegel focused on the impact of Noora’s empathy module. The chatbot first offers a leading statement, such as “I’m feeling really tired lately and it’s been so hard to concentrate,” and then asks the user to assess whether the statement is positive, neutral, or negative. Noora will then grade this response and ask the user to respond empathically to the initial statement. Based on whether the user successfully responds with empathy, Noora either offers a gentle correction or validates a correct response.
The research team carefully crafted prompts with representative examples to ensure that the answers are appropriate. To interact with users, Noora needed to know three things: what kind of statements warrant an empathic response, how to assess whether a user has responded empathetically, and how to offer a user helpful feedback to improve the response if it lacked empathy.
To craft leading statements, the team exposed Noora to “golden” responses and to inappropriate responses. The team both wrote responses themselves and used the LLM to write other responses that they then verified, creating a pool of 330 statements designed to elicit empathetic responses from participants. This means that Noora was never creating leading statements on the fly, which could have potentially led to inappropriate questions.
When it came to responding live to users’ empathetic responses, Noora had freer rein. Taking advantage of the LLM’s abilities for in-context learning, the team simulated users’ personalities and had Noora practice responding to users that showed varying levels of empathy. They also selected difficult cases and provided feedback for Noora to learn from.
An example of the NOORA interface and a sample of an interaction.
Putting Noora to the Test
To see how well Noora stacked up against treatment as usual, Koegel, Lam, and colleagues conducted a randomized trial with 30 participants in which half were assigned to use Noora for four weeks and half received no intervention. Participants using Noora were asked to complete 10 trials per day, five days a week for a total of 200 trials.
Ultimately, Koegel said, the team was looking to evaluate whether Noora could improve users’ empathetic responses and if improvements could be generalized to show empathy in human-to-human communication as well.
“There’s a lot of AI research out there that shows that the ASD students improve using a program, but doesn’t show that it generalizes to real life,” said Koegel. “So that was our main goal.”
Comparing responses from the start of the experiment to the end, Koegel said that 71 percent of participants improved their number of empathetic responses when using Noora.
To see whether this progress could generalize, the team had participants take part in a Zoom call with a team member before and after the experiment, which included leading empathic statements. When reassessed after the intervention, the experimental group scored significantly higher than the control group with an average increase of 38 percent, while the control groups pre- and post-scores were similar. This shows that just four weeks of using the AI program significantly improved verbal empathetic responses.
With this success under Noora’s belt, Koegel and Lam are now interested in testing the effectiveness of other modules as well. They’re also working to open Noora for beta testing for public use and in clinical settings.
Beyond Noora, Koegel said she’s also incorporating AI into other aspects of her autism research as well, including motivational treatment for children with ASD who are in the beginning stages of using communication.
“I’d like to take a lot of the work that I’ve done over the decades that’s face-to-face and see how much we can translate to AI,” said Koegel. “Since kids really like the computer, we want to see if instead of just spending time on their computer or iPhone we can create a learning experience.”
I really don’t think a random D&D table is the place to learn to express empathy. I really wish people would stop acting like local D&D groups are a good way to learn how to socialize in general. I’m not saying you can’t learn things at the table, but the games are not actual reflections of reality and there’s a lot of go along to get along, or just run of the mill toxic group dynamics. The hobby overall can be hard for other minorities to enter, and having a table with someone still learning social skills (especially how to express empathy) and someone from a marginalized group can lead to unfortunate outcomes that your standard DM/group do not have the ability to address. It can lead one or both parties to have negative experiences that reinforce the idea they are unwelcome and leave the rest of the table with negative experiences of playing with ND people or minorities.
Sometimes practicing first with people trained to do this is the best step, and second to that would be practicing empathy in a space where the main goal is bonding rather than another nebulous goal of having fun playing a game. I don’t know if AI is the answer, but trusting your local DM/table to be able to teach empathy is a big ask. It’s almost insulting to the people that teach this and to people with ASD. Teaching empathy can’t be as passive as it is for non-ASD people, and acting like it’s just something they are expected to pick up while also dealing with all these other elements makes it seems like you don’t think it’s something they actually have to work to achieve. I’m not on the spectrum but I have a lot of autistic friends and I would not put just any of them in a D&D situation and expect them and the rest of the table to figure it out.
Also, generally comparing to an unaffected control is the gold standard. They did what is generally needed to show their approach has some kind of effect.
I’ll be honest, I find the framing of the study offensive and I’m not sure if I have the words but I’ll try.
It’s less about this study comparing itself to no intervention instead, but the social & political context of AI being pushed as a way to make care giving more efficient while sacrificing quality.
I don’t personally find the framing offensive, but I’m not on the spectrum so I can’t speak to it from that perspective. My comment was less about the article and more about not offloading that work onto unsuspecting and unprepared people.
That being said, I’m not as anti-ai as maybe some other people might be when it comes to these kinds of tools. The study itself highlights the fact that not everyone has the resources to get the kind of high quality care they need and this might be an option. I agree that sacrificing quality for efficiency is bad, in my post history you can see I made that argument about ai myself, but realistically so many people can potentially benefit from this that would have no alternatives. Additionally, AI will only be getting better, and hopefully you’ve never had a bad experience with a professional, but I can speak from personal experience that quality varies drastically between individuals in the healthcare industry. If this is something that can be offered by public libraries or school systems, so that anyone with the need can take advantage, I think that would be a positive because we’re nowhere near universal physical healthcare, much less universal mental healthcare or actual social development training. I know people who cannot afford healthcare even though they have insurance, so if they were able to go to a specialized ai for an issue I would think it’s a net positive even if it’s not a real doctor. I know that ai is not there yet, and there’s a lot of political and social baggage there, but the reality is people need help and they need it now and they are not getting it. I don’t know how good this ai is, but if the alternative is telling people that are struggling and have no other options that they have to tough it out, I’m willing to at least entertain the idea. For what it’s worth, if I could snap my fingers and give everyone all the help and support they need and it excluded ai, I would choose that option, I just don’t have it. I also don’t know that LLMs really can do this successfully on a large scale, so I would need evidence of that before really supporting it, I just think it shouldn’t be written off completely if it’s showing promise.
It might get cheaper, but that doesn’t mean it’s doing a better job.
if the alternative is telling people that are struggling and have no other options that they have to tough it out
That’s just it, if you’re talking to someone who’s is struggling with this there is already a better option: showing empathy. I suspect our perceived lack of empathy is a reflection of how society treats people in general, we are just more honest about it and recognize it’s mostly platitudes.
By getting better, I mean it will be improving on itself. I never meant to indicate that it will be better than a trained professional.
I agree that showing ND people empathy is the best path forward, but realistically being able to socially signal empathy is a life skill and lacking that skill really only damages their own prospects. It’d be great if it didn’t make people less likely to be employable or less able to build a robust support network, but unfortunately that’s the case. Yes, ASD differences are often a reflection of how society treats people, but a demonstration of empathy is not a platitude. It’s an important way NT and lots of ND connect. If you think that the expression of empathy is difficult for people with ASD because they are more honest, then I think you might be equating lack of empathy with difficulty expressing it. There’s nothing dishonest about saying “I’m sorry that happened to you” unless you are not sorry it happened. It might not be something you would normally verbally express, but if hearing about a bad thing happening to someone doesn’t make you feel for them, then the difficulty isn’t expressing empathy, it’s lacking it. Society certainly does a lot of things for bad or nonsensical reasons, but expressing empathy generally isn’t one of them.
I at no point said that anyone wasn’t worth the time for personal interaction. I said multiple times that my preferred solution would not involve having to resort to AI. That’s such a bad faith interpretation of my position that I can’t imagine this being productive at this point. Best of luck.
The article:
A specialized chatbot named Noora is helping individuals with autism spectrum disorder practice their social skills on demand.
Knowing what to say when co-workers tell you about their weekend is a social skill that many take for granted, but for some individuals with autism spectrum disorder (ASD), this social interaction can be challenging. Struggling to find the correct response in social situations like these can negatively impact social and professional relationships for people with ASD, and can worsen co-occurring conditions like depression.
Research has shown that practicing social interactions with professionals in a clinical face-to-face intervention can improve outcomes for individuals, but these solutions are often costly or not widely available. Lynn Koegel, a clinical professor of psychiatry and behavioral sciences at Stanford University, in collaboration with Professor Monica Lam from Stanford’s Computer Science Department, are the authors of recent research published by the Journal of Autism and Developmental Disorders that investigates the role of AI in filling this gap.
“Our research has shown that face-to-face work does help with social conversation … so we wanted to see if we could translate that into computer use,” said Koegel. “Accessibility is really important because a lot of people don’t have access to a face-to-face provider and the providers can be really expensive.”
Introducing Noora In this work, funded in part by a seed grant from the Stanford Institute for Human-Centered AI and the Kind World Foundation, Koegel and colleagues evaluated interactions between participants with ASD who struggle with social interactions and Noora, a chatbot built with a large language model (LLM).
In one-on-one interactions, which can be written or spoken, Noora offers individualized guidance on a number of social communication scenarios; it helps users learn to ask questions, give compliments, respond empathically, and with other areas of social communication that are often challenging.
In this recent work, Koegel focused on the impact of Noora’s empathy module. The chatbot first offers a leading statement, such as “I’m feeling really tired lately and it’s been so hard to concentrate,” and then asks the user to assess whether the statement is positive, neutral, or negative. Noora will then grade this response and ask the user to respond empathically to the initial statement. Based on whether the user successfully responds with empathy, Noora either offers a gentle correction or validates a correct response.
The research team carefully crafted prompts with representative examples to ensure that the answers are appropriate. To interact with users, Noora needed to know three things: what kind of statements warrant an empathic response, how to assess whether a user has responded empathetically, and how to offer a user helpful feedback to improve the response if it lacked empathy.
To craft leading statements, the team exposed Noora to “golden” responses and to inappropriate responses. The team both wrote responses themselves and used the LLM to write other responses that they then verified, creating a pool of 330 statements designed to elicit empathetic responses from participants. This means that Noora was never creating leading statements on the fly, which could have potentially led to inappropriate questions.
When it came to responding live to users’ empathetic responses, Noora had freer rein. Taking advantage of the LLM’s abilities for in-context learning, the team simulated users’ personalities and had Noora practice responding to users that showed varying levels of empathy. They also selected difficult cases and provided feedback for Noora to learn from.
An example of the NOORA interface and a sample of an interaction.
Putting Noora to the Test To see how well Noora stacked up against treatment as usual, Koegel, Lam, and colleagues conducted a randomized trial with 30 participants in which half were assigned to use Noora for four weeks and half received no intervention. Participants using Noora were asked to complete 10 trials per day, five days a week for a total of 200 trials.
Ultimately, Koegel said, the team was looking to evaluate whether Noora could improve users’ empathetic responses and if improvements could be generalized to show empathy in human-to-human communication as well.
“There’s a lot of AI research out there that shows that the ASD students improve using a program, but doesn’t show that it generalizes to real life,” said Koegel. “So that was our main goal.”
Comparing responses from the start of the experiment to the end, Koegel said that 71 percent of participants improved their number of empathetic responses when using Noora.
To see whether this progress could generalize, the team had participants take part in a Zoom call with a team member before and after the experiment, which included leading empathic statements. When reassessed after the intervention, the experimental group scored significantly higher than the control group with an average increase of 38 percent, while the control groups pre- and post-scores were similar. This shows that just four weeks of using the AI program significantly improved verbal empathetic responses.
With this success under Noora’s belt, Koegel and Lam are now interested in testing the effectiveness of other modules as well. They’re also working to open Noora for beta testing for public use and in clinical settings.
Beyond Noora, Koegel said she’s also incorporating AI into other aspects of her autism research as well, including motivational treatment for children with ASD who are in the beginning stages of using communication.
“I’d like to take a lot of the work that I’ve done over the decades that’s face-to-face and see how much we can translate to AI,” said Koegel. “Since kids really like the computer, we want to see if instead of just spending time on their computer or iPhone we can create a learning experience.”
If only they gave a control group an off-the-shelf social game like LA Noire or a D&D play group
I really don’t think a random D&D table is the place to learn to express empathy. I really wish people would stop acting like local D&D groups are a good way to learn how to socialize in general. I’m not saying you can’t learn things at the table, but the games are not actual reflections of reality and there’s a lot of go along to get along, or just run of the mill toxic group dynamics. The hobby overall can be hard for other minorities to enter, and having a table with someone still learning social skills (especially how to express empathy) and someone from a marginalized group can lead to unfortunate outcomes that your standard DM/group do not have the ability to address. It can lead one or both parties to have negative experiences that reinforce the idea they are unwelcome and leave the rest of the table with negative experiences of playing with ND people or minorities.
Sometimes practicing first with people trained to do this is the best step, and second to that would be practicing empathy in a space where the main goal is bonding rather than another nebulous goal of having fun playing a game. I don’t know if AI is the answer, but trusting your local DM/table to be able to teach empathy is a big ask. It’s almost insulting to the people that teach this and to people with ASD. Teaching empathy can’t be as passive as it is for non-ASD people, and acting like it’s just something they are expected to pick up while also dealing with all these other elements makes it seems like you don’t think it’s something they actually have to work to achieve. I’m not on the spectrum but I have a lot of autistic friends and I would not put just any of them in a D&D situation and expect them and the rest of the table to figure it out.
Also, generally comparing to an unaffected control is the gold standard. They did what is generally needed to show their approach has some kind of effect.
I’ll be honest, I find the framing of the study offensive and I’m not sure if I have the words but I’ll try.
It’s less about this study comparing itself to no intervention instead, but the social & political context of AI being pushed as a way to make care giving more efficient while sacrificing quality.
I don’t personally find the framing offensive, but I’m not on the spectrum so I can’t speak to it from that perspective. My comment was less about the article and more about not offloading that work onto unsuspecting and unprepared people.
That being said, I’m not as anti-ai as maybe some other people might be when it comes to these kinds of tools. The study itself highlights the fact that not everyone has the resources to get the kind of high quality care they need and this might be an option. I agree that sacrificing quality for efficiency is bad, in my post history you can see I made that argument about ai myself, but realistically so many people can potentially benefit from this that would have no alternatives. Additionally, AI will only be getting better, and hopefully you’ve never had a bad experience with a professional, but I can speak from personal experience that quality varies drastically between individuals in the healthcare industry. If this is something that can be offered by public libraries or school systems, so that anyone with the need can take advantage, I think that would be a positive because we’re nowhere near universal physical healthcare, much less universal mental healthcare or actual social development training. I know people who cannot afford healthcare even though they have insurance, so if they were able to go to a specialized ai for an issue I would think it’s a net positive even if it’s not a real doctor. I know that ai is not there yet, and there’s a lot of political and social baggage there, but the reality is people need help and they need it now and they are not getting it. I don’t know how good this ai is, but if the alternative is telling people that are struggling and have no other options that they have to tough it out, I’m willing to at least entertain the idea. For what it’s worth, if I could snap my fingers and give everyone all the help and support they need and it excluded ai, I would choose that option, I just don’t have it. I also don’t know that LLMs really can do this successfully on a large scale, so I would need evidence of that before really supporting it, I just think it shouldn’t be written off completely if it’s showing promise.
It might get cheaper, but that doesn’t mean it’s doing a better job.
That’s just it, if you’re talking to someone who’s is struggling with this there is already a better option: showing empathy. I suspect our perceived lack of empathy is a reflection of how society treats people in general, we are just more honest about it and recognize it’s mostly platitudes.
By getting better, I mean it will be improving on itself. I never meant to indicate that it will be better than a trained professional.
I agree that showing ND people empathy is the best path forward, but realistically being able to socially signal empathy is a life skill and lacking that skill really only damages their own prospects. It’d be great if it didn’t make people less likely to be employable or less able to build a robust support network, but unfortunately that’s the case. Yes, ASD differences are often a reflection of how society treats people, but a demonstration of empathy is not a platitude. It’s an important way NT and lots of ND connect. If you think that the expression of empathy is difficult for people with ASD because they are more honest, then I think you might be equating lack of empathy with difficulty expressing it. There’s nothing dishonest about saying “I’m sorry that happened to you” unless you are not sorry it happened. It might not be something you would normally verbally express, but if hearing about a bad thing happening to someone doesn’t make you feel for them, then the difficulty isn’t expressing empathy, it’s lacking it. Society certainly does a lot of things for bad or nonsensical reasons, but expressing empathy generally isn’t one of them.
Saying that you’re not worth the time for personal interactions but here’s a reason that’s okay is a platitude.
I at no point said that anyone wasn’t worth the time for personal interaction. I said multiple times that my preferred solution would not involve having to resort to AI. That’s such a bad faith interpretation of my position that I can’t imagine this being productive at this point. Best of luck.