AI-Driven Behavioral Support for Neuropsychiatric Disorders

Introduction

Mental health disorders are highly prevalent and many times undertreated worldwide, which drives interest in innovative support strategies. Artificial intelligence (AI)–powered tools have emerged as promising tools to traditional care. They can offer improved capabilities to monitor patients continuously, personalize treatments and help to assess a greater number of patients, even in remote areas. In particular, the AI’s ability to process data from wearables, smartphones and other devices with efficiency can provide a great value for diagnosis, behavioral tracking and treatment of mental disorders.

This is the first whitepaper and we will provide a comprehensive overview of how real-world and emerging AI technologies are supporting individuals with seven common neuropsychiatric conditions. These are Autism Spectrum Disorder (ASD), Attention-Deficit/Hyperactivity Disorder (ADHD) and Depression.

We will focus on applications in behavioral tracking (e.g., monitoring symptoms or activity patterns) and assistive technologies (that is, tools that help manage or treat symptoms), all of which highlight current solutions and cutting-edge research. 

 

AI in Autism Spectrum Disorder (ASD)

Current Applications: 

ASD is characterized by challenges in communication, social interaction and repetitive behaviors. AI-driven assistive technologies are being applied to help individuals with ASD develop social and communication skills and to monitor behavioral states. Notably, socially-assistive robots have been introduced in therapy and educational settings to engage children with ASD in interactive exercises. These robots (such as humanoid robots) can stimulate social cues and also help individuals to have non-judgmental companions to practice interactions. For example, Milo by Robokind is a robot designed to teach emotional understanding and social skills to autistic children. Early studies suggest that children with ASD, in many cases, find these robots engaging and may maintain better focus with a robot tutor than with human therapists in certain tasks. Likewise, smart glasses leveraging AI-based facial expression recognition have shown promise in improving socioemotional skills. One system worth mentioning is the Brain Power Autism System (BPAS) using Google Glass. This system brings real-time cues about others’ emotions during social interactions. In pilot trials, AI-driven smart glasses were well-tolerated by students with ASD and helped them recognize facial expressions and social cues more accurately. Additionally, AI has been integrated into augmentative and alternative communication (AAC) apps for non-verbal individuals – for instance, by embedding a conversational agent that can predict or suggest words or phrases to sharpen the communication. Therefore, tablets and apps can help users to express their needs or feelings better when human communication partners are not immediately available.

Emerging Solutions:

Research on AI for ASD continues to advance rapidly. For example, machine learning algorithms are being developed to detect patterns in behavior and physiology that signal stress or discomfort in non-verbal autistic individuals. With this information, caregivers can assist more proactively. When it comes to experimental wearable sensors (like wristbands or clothing with integrated monitors coupled with AI) can track physiological indicators like heart rate and skin conductance. If this occurs, anxiety or sensory overload can alert a caregiver by sending a trigger automatically. 

Computer vision is also being explored for behavioral tracking, for example using cameras with AI that guide repetitive movements or atypical gaze patterns in autism. These systems could sense behaviors quantitatively, like stimming frequency or eye contact duration, to support behavioral therapy plans and objectively track progress over time. 

Another technological frontier are the robotic therapy assistants, which help adapt to the user’s profile. Researchers are equipping robots with reinforcement learning algorithms so that the robot’s feedback and challenge level, like in a game to practice social turn-taking, can adjust dynamically to the child’s engagement and skill mastery. These personalized robotics interventions are still largely in pilot stages but show potential for improving generalization of social skills beyond the training sessions. 

As we can see, ASD care is witnessing “exciting developments… at the intersection of AI and robotics, as well as in the creation of wearable automated devices like smart glasses,” which “offer substantial potential for enhancing communication, interaction, and social engagement for individuals with autism”. Moving forward, an emphasis on user-centric design for these tools to meet the diverse sensory and cognitive needs in autism and rigorous clinical validation will be critical to translate these innovations into mainstream autism care.

 

AI in Attention-Deficit/Hyperactivity Disorder (ADHD)

Current Applications:

ADHD is marked by inattention, hyperactivity, and impulsivity, and it impacts on daily activities. AI-powered tools for ADHD primarily aim to improve attention regulation, offer cognitive training, and monitor activity levels in an engaging way. A landmark example is EndeavorRx, the first FDA-cleared prescription video game therapy for pediatric ADHD. EndeavorRx uses adaptive algorithms to adjust game difficulty in real-time as the child plays and targets the neural systems involved in attention. Clinical trials demonstrated that playing the game regularly led to improvements in objective attention performance in children with ADHD, which supports its clearance as a digital therapeutic. 

In practice, this game provides a fun, at-home intervention that can complement medication and therapy. Beyond games, wearable devices are used to assist children in staying on task. For instance, the Revibe smartwatch uses motion sensors and an AI-based pattern recognition to detect when a student’s attention wanders, for example reduced purposeful movements or task engagement. It then delivers gentle vibration reminders to refocus. 

 

In a large pilot with over 700 children, such a wearable system was reported to increase attention span by an average of 25 minutes and improve on-task behavior by ~19%, by continuously nudging the child back to the task when distractions were detected. Also, smartphone apps serve as AI-driven organizational coaches for adolescents and adults with ADHD. These apps can prioritize smart to-do lists, send reminder notifications and even use natural language processing to parse task-related emails or messages and create automated schedules. 

Some commercially available apps use AI to learn which reminders or reward structures work best for an individual, for instance varying the timing of instructions or using gamified rewards for task completion based on user response patterns. Although robust clinical trial data on these apps are still limited, user feedback is positive regarding improved time management and reduced stress in daily task handling. Importantly, many of these tools also serve a tracking function. In other words, they manage data on the user’s focus duration, fidgeting, routine adherence, etc., which can be valuable for clinicians. Objective data from wearables and apps can complement parent and teacher questionnaires in measuring ADHD symptom patterns across settings.

 

Emerging Solutions:

On the research front, AI is being leveraged to make ADHD interventions more personalized and predictive. One area of development is machine learning models that analyze multi-modal data like cognitive test performance, keyboard use patterns and physical activity from sensors. These data help to simplify ADHD diagnosis and to adapt individual treatment. 

Early studies suggest that supervised learning algorithms can identify unique digital activity “signatures” of ADHD, for example characteristic pauses and errors in computerized attention tasks or distinct motor activity profiles during the day. These types of models could eventually support clinical professionals by providing an objective assessment to augment clinical evaluations.

Another emerging application is virtual reality: immersive VR environments combined with AI are being tested to train focus and self-control. A VR training program could present distractions in a virtual classroom and use eye-tracking plus AI to measure the user’s attention; it can dynamically affect the environment difficulty with more or fewer distractions. All of this in response to the user’s performance to strengthen the attention control progressively. 

Preliminary results from these VR-based interventions, like the NeuroTrainer, indicate potential improvements in users’ ability to ignore distractions, though more research is needed. Researchers are also investigating predictive analytics in ADHD management, for example using AI to analyze patterns of behavior or even speech, in verbal children, to predict when a patient is likely to have a particularly unfocused day or an impulsive episode. 

If reliable triggers or early warning signs can be identified, such as poor sleep the night before and is detected via a sleep tracker, the next-day hyperactivity can be correlating. These apps could proactively adjust the day’s schedule or alert caregivers to implement coping strategies. 

AI applications in ADHD range from assistive tech that engages and guides patients to analytical tools that support objective monitoring and personalization of care. The technology has a huge impact and many are already in real-world use with evidence of benefit. The ongoing innovations hold great promise and will be improved even further, including self-management and symptom tracking for ADHD patients.

 

AI in Depression

Current Applications: 

Depression is a widespread mood disorder characterized by persistent sadness, loss of interest and cognitive and physical symptoms. Because it is often a chronic and relapsing condition, continuous support and monitoring can improve outcomes significantly. This is where AI tools have made a considerable impact. One prominent category is are AI-driven conversational agents, also called chatbots, for mental health support. Applications like Woebot Health, Wysa and Youper use natural language processing and therapeutic frameworks, such as cognitive-behavioral therapy techniques to engage users in chat conversations. These chatbots are available 24/7 on smartphones and can help users reframe negative thoughts, practice coping skills or simply provide a nonjudgmental space to have conversations. 

It is important to say that clinical evaluations have shown promising results. In one trial, college students with depression and anxiety who interacted with a cognitive behavioral therapy (CBT)-based chatbot for several weeks reported significantly lower depressive symptoms compared to a control group. The findings suggest that well-designed AI chatbots can reduce mild-to-moderate depression symptoms and act as scalable and accessible therapeutic tools. These tools are currently available direct-to-consumer and some healthcare systems are beginning to offer them as a supported self-care option.

Another area where AI is supporting depression management is behavioral tracking through smartphones and wearable devices. Researchers and digital health startups have developed systems to monitor indicators of mood and functioning passively. This is often referred to as digital phenotyping. For example, a smartphone can collect data continuously on a person’s activity levels (via accelerometer), sleep patterns (via phone use at night or a connected smartwatch), social interaction (frequency of calls or texts) and even voice tone during calls.

Machine learning models process these data streams to detect patterns associated with depression. A reduction in mobility and social interaction, combined with late-night phone usage, might signal a downturn in mood. Some modern smartphone apps for depression already use simple AI algorithms to alert users or caregivers when certain risk patterns, like prolonged inactivity or staying at home for days are detected. This signals a suggested check-in or coping strategy. Wearable sensors add another dimension, for instance changes in heart rate variability or skin conductance can point to emotional distress in an individual. 

Pilot studies have demonstrated that AI algorithms using wearable data can sometimes predict worsening depression before the individual recognizes it, enabling earlier intervention. While these passive monitoring approaches are mostly in experimental or early deployment phases, they hold significant promise for transforming depression care from a reactive to a proactive model.

In addition to tracking and chatbots, AI is also being used in personalized treatment planning for depression. In research settings, algorithms analyze patient characteristics and responses to past treatments to recommend optimal medication or therapy choices, which is an example of AI-driven clinical decision support. Some healthtech companies are developing AI tools to assist psychiatrists by predicting which patients might respond best to antidepressant A vs. B, based on large datasets of prior patients. However, this application is still emerging and requires regulatory approval for clinical use. 

Another application is robots. Robotics has seen niche use-cases such as social companion robots for individuals with depression or dementia-related depression. For example, the robot PARO, a plush baby seal with AI behaviors, has been used in elderly care settings to reduce loneliness and improve the person’s mood. While not a treatment right now, these robots use simple AI to respond to touch and sound, and provide comfort and stimulation. This can alleviate depressive feelings in some users.

Emerging Solutions:

Looking ahead, more sophisticated AI systems are being investigated to deepen our understanding and treatment of depression. Voice analysis AI is one promising tool. Subtle changes in a person’s voice such as flatter intonation, slower speech or longer pauses can all be indications of depressive mood. AI models trained on voice recordings have shown ability to distinguish individuals with depression from healthy controls with notable accuracy by detecting vocal biomarkers. In the future, a short voice note or phone call could be analyzed to gauge a patient’s mental state and complement clinical assessments. 

Similarly, computer vision algorithms are being tested to detect facial expressions of emotion. A front camera on a smartphone or computer could potentially monitor facial affect during the day and give objective data on emotional blunting or mood swings. Another cutting-edge area is generative AI for therapy, where large language models are being customized to serve as more advanced therapeutic chatbots. The purpose of these systems is to engage in deeper conversations, understand context over long dialogues and even generate therapeutic exercises dynamically. 

Early prototypes have shown they can hold empathetic dialogues and provide psychoeducation, but keeping their safety and efficacy, so they do not produce harmful or insensitive responses, is an active area of research. Lastly, the integration of multi-modal data (which combines voice, text, activity and physiological signals) in AI models is an emerging approach to create a holistic “mood radar” for depression. By placing these signals together, AI might achieve more reliable detection of mood shifts than any single source alone. These integrated systems are under evaluation in clinical trials and could form the basis of next-generation depression management platforms to alert care teams of early warning signs and provide individuals with proper support or resources.

 

Synthesis and Way Forward

Across ASD, ADHD and Depression we have seen that AI already performs three clinically valuable functions: (1) continuous detection of symptom shifts via multisensor data, (2) adaptive and engaging digital interventions that extend care beyond the clinic, and (3) objective outcome measurement to guide precision treatment. Early evidence, from an FDA-cleared videogame for ADHD to RCT-validated CBT chatbots for depression shows that well-designed AI tools can improve attention, mood and social cognition at scale.

Yet deployment at population level will require:

1. Interoperable data standards so that sensor streams can be shared, with consent, across apps, electronic health records and research cohorts.
2. Hybrid-care protocols that integrate AI output in existing workflows to preserve the clinician’s central role and, at the same time, decrease caseloads.
3. Rigorous real-world evidence (RWE), including post-market algorithm auditing, to maintain equity, transparency and sustained clinical benefit.
4. Value-based reimbursement that rewards outcomes rather than software licensing alone and stimulates continuous improvement.

If these aspects are put in place, AI can shift care for ASD, ADHD and depression from episodic, resource-intensive interventions to continuous, personalized and preventive support. The three conditions described in this whitepaper already account for a substantial share of global mental-health burden. Scalable AI solutions could therefore unlock outsized public-health impact in the near and long term.

 

References

1. Akili Interactive Labs. (n.d.). Discover EndeavorRx. Retrieved May 8, 2025, from https://www.hcpendeavorrx.com/discover-endeavorrx/
2. Iannone, A., & Giansanti, D. (2024). Breaking barriers—The intersection of AI and assistive technology in autism care: A narrative review. Journal of Personalized Medicine, 14(1), 41. https://doi.org/10.3390/jpm14010041 
3. Jiang, Z., Chan, A. Y. L., Lum, D., Wong, K. H. T. W., Leung, J. C. N., Ip, P., Coghill, D., Wong, R. S. M., Ngai, E. C. H., & Wong, I. C. K. (2024). Wearable signals for diagnosing attention-deficit/hyperactivity disorder in adolescents: A feasibility study. JAACAP Open. Advance online publication. https://doi.org/10.1016/j.jaacop.2024.11.003 
4. Keshav, N. U., Salisbury, J. P., Vahabzadeh, A., & Sahin, N. T. (2017). Social communication coaching smartglasses: Well tolerated in a diverse sample of children and adults with autism. JMIR mHealth and uHealth, 5(9), e140. https://doi.org/10.2196/mhealth.8534 
5. Kim, W. P., Kim, H. J., Pack, S. P., Lim, J. H., Cho, C.-H., & Lee, H.-J. (2023). Machine learning–based prediction of attention-deficit/hyperactivity disorder and sleep problems with wearable data in children. JAMA Network Open, 6(3), e233502. https://doi.org/10.1001/jamanetworkopen.2023.3502 
6. Mattingly, G. (2024, October 30). Gamified interventions for ADHD: Wearables, digital therapeutics. ADDitude Magazine. https://www.additudemag.com/digital-therapeutics-adhd-lumosity-meta-quest-endeavorrx/
7. Robokind. Advanced social robots. Retrieved May 8, 2025, from https://www.robokind.com/advanced-social-robots
8. Terhorst, Y., Messner, E.-M., Opoku Asare, K., Montag, C., Kannen, C., & Baumeister, H. (2025). Investigating smartphone-based sensing features for depression severity prediction: Observation study. Journal of Medical Internet Research, 27, e55308. https://doi.org/10.2196/55308
9. U.S. Food and Drug Administration. (2020, June 15). FDA permits marketing of first game-based digital therapeutic to improve attention function in children with ADHD. https://www.fda.gov/news-events/press-announcements/fda-permits-marketing-first-game-based-digital-therapeutic-improve-attention-function-children-adhd