By Sharon Mozgai, Director, Virtual Human Therapeutics Lab, ICT
Sharon Mozgai, Director, Virtual Human Therapeutics Lab, joined ICT in 2015 as a Research Analyst in the Virtual Humans Group, then served as a Lead R&D Scientist, before becoming Associate Director of MedicalVR. There, she drew on psychology, medicine, neuroscience, physical and occupational therapy, to evaluate where Virtual Reality (VR) could add value over traditional assessment and clinical intervention approaches. Here she explores the therapeutic potential of AI-powered interactions between bio-based and virtual humans in future healing modalities.
One of the most significant advances virtual humans and AI have brought to virtual therapeutics is the development of empathic interfaces. Advanced AI models enable virtual humans to generate empathic responses and provide compassionate support and validation that fosters rapport and trust. Advances in emotion recognition that allow virtual humans to interpret an individual’s emotional state based on both verbal and non-verbal cues, such as tone of voice and facial expressions, along with advances in natural language processing, contribute to trust and engagement in virtual therapeutics interventions.
Virtual humans will revolutionize digital therapeutics by offering scalable, personalized interventions, learning from interactions, and providing invaluable insights for enhanced user engagement and treatment outcomes. AI-powered algorithms will enable virtual humans to deliver personalized and adaptive interventions to a large number of users simultaneously, addressing scalability challenges often encountered in traditional therapeutic approaches. By leveraging AI, virtual humans will be able to continuously learn from user interactions and feedback, which will improve their effectiveness over time and enhance user engagement. AI will also facilitate data-driven insights and analytics, providing valuable information about user behavior, preferences, and treatment outcomes.
Joining ICT
In 2015, I moved to Los Angeles, and a former professor, noting my background in psychology and computational linguistics, urged me to explore the groundbreaking research being conducted at ICT. It was through this recommendation that I crossed paths with Jon Gratch and was invited to become a member of his research team.
My initiation into the realm of virtual humans began with the role of the Wizard-Of-Oz, orchestrating the interactions from behind the scenes by controlling the thousands of possible utterances of his group’s Negotiation Agent in a large research study.
The process of building virtual human experiences quickly captured my interest, reminding me of the camaraderie of the team sports I had played in the past. Constructing a successful virtual human demands a diverse array of skills, including those of computer scientists, linguists, technical artists, social scientists and more.
Collaborating within such a multidisciplinary team to craft clinically meaningful interactions became my driving passion, an experience I had firsthand under the mentorship of Skip Rizzo and the MedVR lab at ICT.
In 2023, I was promoted to Director of my own laboratory within this domain: VHTL, the Virtual Human Therapeutics Lab. The primary objectives of VHTL encompass:
- Pioneering novel methodologies for therapeutic interventions leveraging virtual human technology.
- Exploring the integration of advanced AI algorithms to enhance the realism and efficacy of virtual human interactions.
- Collaborating with healthcare professionals to validate and implement virtual human-based therapies in clinical settings, striving to enhance patient access to care, outcomes, and well-being.
- Prioritizing ethical considerations in all stages of virtual human development, from design to implementation, ensuring that AI-driven interventions uphold principles of fairness, transparency, accountability, and user privacy.
- Engaging in interdisciplinary discourse and collaborating with ethicists, policymakers, and stakeholders to establish guidelines and best practices for the responsible use of AI in healthcare settings.
Augmenting Patient Care
Virtual humans can augment traditional methods by offering additional support and resources to patients. Virtual humans are available 24/7, which can ensure timely access to information and guidance, complementing the care already provided by healthcare professionals.
Through personalized interactions, virtual humans can address patient concerns and provide educational materials tailored to individual needs, enhancing patient understanding and engagement in their care. Virtual humans offer a scalable solution, extending the reach of healthcare services to a larger population with the goal of maintaining consistency and quality of care. Virtual humans can also help combat geographical barriers to care.
Ultimately, by integrating virtual humans into traditional healthcare settings, patients can benefit from a holistic approach that combines the expertise of healthcare professionals with the accessibility and support provided by digital technology.
Considering Ethics
When designing virtual humans for therapeutic purposes, it’s crucial to prioritize safety, security, and accessibility from the outset, particularly around how and where sensitive information is stored within these platforms.
While recent advances in AI are exciting, a core value of our lab is to responsibly integrate these technologies into our virtual humans. For example, Large Language Models (LLMs) like chatGPT have shown incredible capabilities in natural language understanding and generation, but there are still challenges in ensuring accuracy and context-aware responses.
The cost of making mistakes, especially in sensitive areas like mental health support, can have significant consequences for users. Developers of virtual humans should also actively audit the datasets used to train models in order to identify biased or limited representations of certain demographic groups, which can result in biased responses or recommendations.
Additionally, virtual human interfaces should be designed in a way that considers the needs of individuals who may have difficulties processing visual or auditory information, ensuring inclusivity for all users.
Future Developments
One of the most exciting visions of the future includes the evolution of Just-In-Time Adaptive Interventions (JITAIs) that leverage smartphone and wearable technology. JITAI is an intervention design that aims to provide the right support at just the right time, by adapting to an individual’s changing state. Smartphones and wearables offer a novel opportunity for data collection through passive sensing that includes voice and speech analysis, activity tracking, sleep patterns, social media engagement, and biometric sensing, combined with contextual data analysis and machine learning algorithms to provide insight into an individual’s mental and physical state without requiring active input from the end-user.
Our lab envisions a future in which JITAIs have the potential to revolutionize personalized support and assistance in real-time, empowering virtual humans to reach out to individuals at their moments of greatest need.
Virtual humans hold immense promise in optimizing patient outcomes and enriching the healthcare landscape. There is little doubt that virtual humans will become ubiquitous in healthcare across multiple domains. Many of our lab’s applications are focused on delivering personalized healthcare experiences and therapeutics interventions at-scale. Virtual humans will also revolutionize patient engagement, clinical decision making and care-giver support.
It would be a missed opportunity if there was no collaboration between academia, industry and healthcare practitioners to address the technical challenges and ethical considerations in the development of virtual humans. Industry partners have the resources, funding and infrastructure to bring virtual humans out of the lab and into the hands of users on a large scale, helping to drive the adoption of these technologies into everyday therapeutic practices.
In an ideal world, companies that specialize in AI collaborate with academia and healthcare practitioners to translate research findings into evidence-based applications with clinical relevance. Academics and healthcare practitioners also have an important role to play in the ethical review and oversight of assessing bias in AI applications. Industry should actively engage academics and healthcare professionals to identify potential sources of bias in the underlying algorithms and decision making processes of virtual humans. Collaborations across these three domains will improve patient outcomes and foster trust in AI-enabled therapeutic interventions.
//
BIO:
Sharon Mozgai is the Director, Virtual Human Therapeutics Lab (VHTL), at ICT. She holds a Masters in Human Development and Psychology from Harvard University, a B.A. in Psychology from New York University (NYU), and has worked as a Research Associate at Harvard Business School (Organizational Behavior Unit), and as a Text Analyst at Massachusetts Institute of Technology (MIT) on a DARPA-funded natural language processing (NLP) study. She joined ICT in 2015 as a Research Analyst in the Virtual Humans Group, then served as a Lead R&D Scientist, before becoming Associate Director of MedicalVR, drawing on psychology, medicine, neuroscience, physical and occupational therapy, to evaluate where Virtual Reality (VR) could add value over traditional assessment and clinical intervention approaches.