Prof. Joseph Levine

Talk 79: Exploring the Integration of AI-Generated Avatars into Reference Group Focused Therapy (RGFT)

By Prof. Levine & Dr. Salganik

Hello readers,

The field of psychotherapy is constantly evolving, with an increasing emphasis on innovative approaches to addressing the multifaceted nature of mental health challenges. Among these developments, Reference Group Focused Therapy (RGFT), originally developed by one of us [I.S.], stands out as a contribution to short-term interventions based on social psychology.

RGFT distinguishes itself by taking a comprehensive view of an individual's social environment, which includes both external reference groups such as family and workplace, and internalized representations of significant others. This holistic perspective recognizes that an individual's thoughts, feelings, and behaviors are profoundly shaped by their interactions and identifications within these social contexts.

It is assumed that understanding and addressing the dynamics within the person’s social matrix is essential for therapeutic progress. This fundamental principle creates a compelling rationale for exploring tools that can effectively represent and interact with this complex social environment in the therapeutic setting.

At the same time, the application of artificial intelligence (AI) in mental health heralds growth in the development and use of AI-generated avatars in various therapeutic contexts.

The potential application of AI avatars in areas such as the treatment of auditory hallucinations in people with psychosis demonstrates the transformative potential of digital representations to alter deeply internalized experiences.

We hypothesize that the success of these interventions suggests that similar approaches may be effectively adapted to represent the internalized significant others that are central to the principles of RGFT.

Below we will delve into how AI-generated avatars of significant others can be effectively and ethically integrated into RGFT to improve therapeutic outcomes. By examining the theoretical foundations of RGFT and the current landscape of AI avatars in psychotherapy, we seek to provide an understanding of the possibilities, benefits, challenges, and ethical considerations associated with such an innovative integration.

Understanding Reference Group Focused Therapy (RGFT)

Reference Group Focused Therapy (RGFT) represents a new, short-term, integrative psychotherapeutic approach rooted in social psychology.

A defining characteristic of RGFT is its comprehensive consideration of the individual's entire social environment, which includes not only his external reference groups, such as family, workplace, and friends, but also his internalized groups, which are mental representations of significant others, alongside his subcultural background.

This broad perspective distinguishes RGFT from many other psychotherapies that may focus primarily on internal psychological processes without explicitly addressing the pervasive influence of the social context.

RGFT was designed as a client-centered psychotherapy intended for short-term treatment in the treatment of a variety of clinical entities, including phobias, obsessive-compulsive disorder (OCD), eating disorders, adjustment disorder, somatization, anxiety disorder, and panic attacks, and others.

The theoretical foundations of RGFT are built on a number of established social psychological theories, as well as social learning theory, object relations theory, attachment theory, and other related approaches. The importance of the social environment in shaping an individual's inner world and behavior is a central tenet of RGFT.

The concept of "reference group" itself was first formulated by Hyman, who studied how people evaluate themselves by comparing themselves to specific groups and people they use as a frame of reference.

RGFT also incorporates principles of social learning theory, recognizing that people learn through observing, modeling, and imitating the behaviors and attitudes of others. Furthermore, the treatment also draws on object relations theory, which emphasizes the significance of internalized representations of significant others and how these internal relational dynamics influence current functioning.

Attachment theory also plays a role in RGFT, recognizing the profound influence of early attachment experiences with primary caregivers [usually parents] on the development of internalized reference groups and subsequent relationship patterns.

Several core concepts are central to the practice of RGFT. These include a recognition of external reference groups, which are the real-world social groups to which the individual belongs or interacts.

Equally important are internalized reference groups, which are the mental representations of significant others that a person carries within them. These internalized figures can have a powerful and often unconscious influence on a person's thoughts, feelings, and behaviors, acting almost like internal "software."

RGFT also takes into account the individual's subcultural background, and recognizes the broader social and cultural influences that shape their experiences and perspectives.

During the assessment phase of RGFT, a specific tool called Trigger Event Analysis is often used. This tool helps to understand the individual’s unique vulnerabilities and allows for the prediction of potential future conflicts, as well as initiating and identifying more effective strategies to avoid or minimize them.

The main goals of RGFT are to bring these unconscious social influences into awareness, to enable a deeper understanding of internalized significant figures and their influence, to resolve conflicts that may arise between internal and external reference groups, and ultimately to promote the patient's process of autonomy along with the improvement of their subjective distress (2).

Artificial Intelligence-Generated Avatars in Psychotherapy: The Current State of Affairs

In the emerging field of e-mental health interventions, the is a gradual increase in the use of avatars. Avatars, defined as digital self-representations, allow people to interact in computer-based virtual environments. This technology is increasingly finding applications in various therapeutic modalities.

One notable application is in the treatment of auditory hallucinations experienced by people with psychosis, a therapeutic approach known as avatar therapy (4).

Avatar therapy involves the patient who hears voices collaborating with a therapist to create a personalized computer-generated avatar that visually and aurally represents the primary disturbing sound the patient is experiencing.

Therapy sessions involve a three-way dialogue between the voice-hearing patient, the therapist (who speaks as themselves and operates the avatar), and the avatar. The goal is to help the voice-hearer confront the voice, challenge their negative statements, and ultimately take control. Studies have shown that avatar therapy can lead to a significant reduction in the frequency and distress associated with the voices.

AI-generated avatars are also being used to facilitate social skills training, especially for people with autism spectrum disorder. In these applications, avatars are used to simulate various social situations in a safe and controlled virtual environment.

Participants can practice social interactions, such as participating in conversations or navigating different social scenarios with avatars that can be programmed to exhibit specific behaviors and responses. This allows people to develop and hone their social skills in an environment where they can experiment without the anxieties associated with real-world social encounters.

Furthermore, avatars have shown promise in reducing social anxiety and communication barriers in therapeutic settings. The use of avatars can provide a sense of anonymity and psychological distance for people experiencing social anxiety, which may make it easier for them to engage in therapy and express inner thoughts and feelings. This may foster a greater willingness to disclose personal information and participate more actively in the therapeutic process.

AI-generated avatars are also playing an increasingly important role in remote care delivery. In virtual care settings, avatars can create more immersive and interactive experiences for patients, improving their engagement and strengthening the therapeutic alliance. This is particularly beneficial for individuals who may face geographic barriers or mobility constraints that prevent them from attending in-person care sessions.

Beyond direct patient care, AI avatars are also being used as tools for training therapists. AI-powered avatars can simulate patients with a wide range of conditions and personality types, providing a safe and controlled environment for therapists to practice their clinical skills, receive immediate, individualized feedback, and develop improved readiness for real-world clinical encounters.

The technology underlying AI therapy avatars is advancing rapidly, combining sophisticated AI algorithms, including machine learning and natural language processing, alongside augmented reality technologies such as virtual reality (VR) and augmented reality (AR) (6).

These technologies enable the creation of responsive and interactive virtual characters that are capable of simulating human-like interactions, understanding and responding to users’ questions, emotions and behaviors in real time. VR and AR technologies further enhance the therapeutic experience by creating immersive and interactive environments that can augment traditional therapy or provide entirely virtual therapeutic spaces.

Integrating AI Avatars in RGFT

The principles and techniques of reference group-focused therapy present several compelling opportunities for integration with AI-generated avatars of significant others. By leveraging AI’s capabilities to create personalized, interactive digital representations, RGFT can enhance its effectiveness in addressing the complex social dynamics that impact an individual’s mental well-being.

One potential avenue for integration involves using avatars to represent specific individuals from the patient's external reference groups in individual or group therapy. For example, in an individual or group session, patients could interact with avatars that represent their family members, partners, or colleagues who are currently contributing to interpersonal difficulties.

This will provide a controlled and safe environment to explore with the individual or group therapist the dynamics of these relationships, practice different communication styles, and perhaps work on conflict resolution with the therapists or group without the immediate pressures and potential negative consequences of interacting with actual people.

Furthermore, AI avatars can be essential in representing a patient's internalized representations of significant others, even if those people are from the past or no longer physically present. RGFT highlights the profound impact of these internalized figures on current functioning.

By creating avatars that embody the perceived characteristics, behaviors, and even the “voice” of these introverted characters, patients could engage in a more direct and explicit exploration of their influence. This can be particularly valuable for working through unresolved issues, processing past experiences, and understanding the ongoing impact of these internalized relationships.

Given RGFT's focus on "internalized figures’ board" and the concept of the "leader self" in figures board’ management, AI avatars can be used to illustrate these different aspects or "voices" within the patient's internalized reference group.

Each significant internalized figure can be represented by a distinct avatar, allowing the patient and therapist to observe and facilitate dialogue with these internal figures.

This visualization can promote greater self-awareness, help identify conflicting internal dynamics, and support the integration process, which is a central goal of RGFT.

The personalization capabilities of AI are essential for effective integration. AI can be used to customize the avatars’ appearance, voice, mannerisms, and even typical expressions based on detailed descriptions of the patient and the experiences of their significant others.

Highly personalized avatars can enhance the sense of realism and emotional connection, making therapeutic interactions feel more authentic and impactful to the patient. The level of detail can be customized based on therapeutic goals and the patient’s comfort level.

Potential Benefits of AI Avatars in RGFT

The integration of AI avatars into RGFT has the potential to offer a number of significant benefits that can enhance the therapeutic process and improve patient outcomes. One key benefit lies in the improved visualization of internal conflicts and relationship dynamics. AI avatars can make the abstract concepts of internalized significant others and their impact on the patient more concrete and observable.

By seeing a digital representation that embodies the characteristics of a specific person or an internalized “voice,” patients may gain a deeper and more immediate understanding of these unconscious dynamics. This visual representation can facilitate insight into how these characters influence their thoughts, feelings, and behaviors in real-world situations. The interactive and innovative nature of AI avatars can also lead to increased patient engagement and motivation in treatment.

For some people, especially those who may find traditional talk therapy less appealing or engaging, the use of technology and virtual representations can make the therapeutic process more interesting and accessible. This increased engagement may improve patient compliance with treatment and lead to more positive outcomes.

AI-assisted avatars provide unique opportunities for safe role-playing with representations of significant others. Patients can practice new communication styles, assertiveness techniques, and coping mechanisms in a controlled and supportive environment. Interacting with an avatar allows them to experience situations that closely mimic different reactions and behaviors without fear of negative real-world consequences that may result from direct interaction with the real person.

This safe space for practice can build confidence and facilitate the transfer of learned skills to real-life relationships. The potential for personalized therapeutic experiences is another significant benefit. Artificial intelligence allows for the creation of personalized avatars that closely resemble the patient’s perception of their significant others, both in appearance and behavior.

AI can be programmed to tailor the avatar’s responses and interactions based on the patient’s history, identified triggers, and specific treatment goals. This level of personalization can make treatment feel more relevant and tailored to the individual’s unique social context and inner world.

Interacting with avatar representations of significant others, especially those related to past trauma or difficult emotions, can facilitate emotional expression and catharsis. For patients who have difficulty expressing certain emotions, such as anger or grief, to the real person or even in their own minds, the avatar can serve as a safe target for these feelings. This emotional release, in the supportive context of therapy, can be a crucial step in processing difficult experiences and moving toward healing.

Observing their responses and behaviors when interacting with the avatars can provide patients with important self-awareness and insight into their interpersonal patterns. By reviewing recordings of their encounters or reflecting on their interactions, patients may notice recurring patterns in their communication style, emotional responses, or defense mechanisms triggered by specific types of interactions with the avatars.

This increased self-understanding is a central goal of RGFT and can empower clients to make positive changes in their relationships.

Finally, AI avatars offer a unique opportunity to address past relationships and internalized figures who are no longer physically present. For individuals grieving the loss of a significant other or needing to work through unresolved issues with someone from their past, interacting with an avatar that embodies that person can provide a powerful means of engagement and processing.

This can be particularly helpful in bringing closure, understanding the ongoing impact of these relationships, and moving forward. Below is a table summarizing the benefits of using AI avatars in psychotherapy in general and RGFT in particular:

Challenges and Considerations

While the integration of AI avatars into RGFT presents many potential benefits, it is also important to recognize and carefully consider the potential challenges and limitations associated with this innovative approach.

One significant challenge lies in the accuracy and authenticity of the avatar representation. Capturing the complex nuances of a real person’s personality, communication style, and emotional expressions, or even the intricate details of an introverted character, is a significant technological and psychological hurdle. If the avatar’s representation is inaccurate or feels inauthentic to the patient, it can hinder the therapeutic process, erode trust, and even reinforce negative beliefs rather than challenge them.

There is also the potential for misinterpretation of avatar behavior and responses. Patients may attribute intentions or emotions to an AI-powered avatar that are inconsistent with intended therapeutic goals or underlying programming. A neutral avatar response, for example, may be misinterpreted as cold or disinterested by a patient seeking empathy, even if the intention was to create a nonjudgmental space. Such misinterpretations can lead to misunderstandings, frustration, or negative emotional reactions.

Technological barriers and accessibility issues also need to be considered. Implementing AI avatar therapy in RGFT may require specialized software, hardware such as VR headsets, and a level of technical expertise that may not be available in all clinical settings or to all therapists. The cost of these technologies and the need for ongoing technical support may also limit the accessibility of this approach, especially in under-resourced settings or for patients with limited financial means.

A crucial consideration is the risk of over-reliance on technology and the potential impact on the development of a genuine human connection in therapy. Psychotherapy relies heavily on the therapeutic relationship and authentic interactions between people. It is essential to ensure that AI avatars serve as a tool to enhance, rather than replace, these essential human elements.

Overreliance on virtual interactions may impair a patient's ability to develop and maintain meaningful connections in their real-life relationships. Incorporating virtual representations into a live group therapy setting may create complexity in terms of engagement, communication flow, and an overall sense of group cohesion (in case of a group therapy).

Patients interacting with avatars may feel disconnected from other group members (in case of a group therapy) who engage in more traditional forms of interaction, which could impact the therapeutic benefits of the group setting. The potential for technical glitches and disruptions during therapy sessions is another practical consideration.

Software crashes, avatar glitches, or connectivity issues can disrupt the flow of care and cause frustration for both patient and therapist. Ensuring the reliability and stability of the technology is essential for a positive therapeutic experience.

Finally, the effective integration of AI avatars into the RGFT process will require rigorous training and guidance of therapists. Therapists will need specific training to understand the capabilities and limitations of AI avatars, how to personalize and manage their interactions, and how to effectively integrate them into RGFT principles and techniques. Without appropriate training, there is a risk of using the technology ineffectively or even in ways that are not beneficial to therapeutic goals.

We also note that the integration of AI-generated avatars into RGFT raises a number of important ethical considerations that must be carefully addressed to ensure responsible and beneficial implementation.

Privacy and data security are paramount concerns. The creation and use of avatars representing significant others will inevitably involve the processing of sensitive personal information provided by the patient. Ensuring the confidentiality of this data, as well as the content of interactions with the avatars, is critical.

Strong security measures and clear privacy protocols must be in place to protect patient information from unauthorized access or breaches. Obtaining informed consent from patients is equally essential. Before using AI avatars in treatment, patients should fully understand the nature of the technology, how it will be used, the potential benefits and risks involved, and their right to refuse or discontinue its use at any time. Therapists should have open and transparent discussions with patients, addressing any questions or concerns they may have to ensure they are making a fully informed decision.

The potential for bias in AI algorithms and avatar representations is another critical ethical consideration. AI algorithms are trained on data, and if that data reflects social biases, the resulting avatars may perpetuate harmful stereotypes or not accurately represent the range of human experiences. This could negatively impact the therapeutic process and undermine the goals of RGFT. Developers and therapists must be vigilant in identifying and mitigating potential biases in the AI systems used to create and animate avatars.

The impact of AI avatars on the authenticity of the therapeutic relationship and group dynamics needs careful consideration. While avatars can offer benefits, it is important to ensure that their use does not diminish the real human connection that is fundamental to effective psychotherapy. Therapists should be aware of how the introduction of AI entities may change the dynamics within therapy and strive to maintain a focus on authentic interpersonal engagement.

Defining boundaries and responsibilities when interactions involve AI entities is also critical. Clear guidelines should be established regarding the therapist’s role in moderating interactions with avatars, addressing any unexpected behaviors or responses from the AI, and managing the client’s emotional responses to these interactions. Therapists should be prepared to intervene and guide the process effectively.

Finally, ethical considerations related to the emotional impact of interacting with AI representations of significant others must be considered, especially in cases of trauma or loss. While avatars can be a powerful tool for processing difficult experiences, interacting with a digital representation of a deceased loved one or an abusive figure can be emotionally challenging or even re-traumatizing for some patients. Therapists should carefully assess the patient’s readiness and emotional stability before using avatars in such sensitive contexts and be prepared to provide appropriate support and inclusion.

Finally, the integration of AI-generated avatars into peer-centered therapy is an emerging area with significant potential for future research.

Several key questions and directions for inquiry warrant further investigation to understand the efficacy, optimal implementation, and ethical implications of this integrated approach.

Pilot studies are needed to assess the feasibility and acceptability of using AI avatars representing significant others in RGFT sessions. Such research could explore patient and therapist experiences, identify potential implementation challenges, and gather initial data on the impact of avatar interactions on therapeutic processes.

Comparative studies can be conducted to directly compare the outcomes of RGFT delivered with and without the integration of AI avatars. These studies can examine various measures of therapeutic progress, such as symptom reduction, improved interpersonal functioning, and increased self-awareness. Research is also needed to determine the optimal design and customization strategies for AI avatars in the context of RGFT. Investigating how different levels of avatar realism, behavioral customization, and interactive capabilities impact patient engagement and treatment outcomes.

The impact of avatar interactions on the therapeutic alliance between patient and therapist should also be investigated, as well as in group therapy on overall group cohesion in RGFT. Further research is also needed to delve deeper into the ethical implications of using AI avatars in this specific therapeutic context. This includes examining issues related to privacy, informed consent, potential biases in avatar representations, and the long-term impact of interacting with AI entities on patient well-being.

Developing and evaluating therapist training protocols for the effective integration of AI avatars into RGFT practice is critical. Research could focus on identifying the key skills required for therapists to use this technology effectively and developing training methods to equip them with these skills.

Studies could also explore the use of AI avatars to represent specific types of internalized characters or to address specific clinical challenges within RGFT. For example, research could explore the effectiveness of using avatars to work with critical internalized parents or to address difficulties in specific types of relationships.

The potential for combining AI avatars with other emerging technologies, such as virtual reality (VR) or augmented reality (AR), to further enhance the therapeutic experience in RGFT also warrants exploration. Immersive virtual environments or augmented reality overlays could create even more engaging and impactful therapeutic interactions.

Finally, longitudinal studies are needed to assess the long-term impact of AI avatar-assisted RGFT on patient outcomes and to understand the potential lasting effects of this therapeutic approach. In conclusion, the integration of AI-generated avatars of significant others in RGFT presents a compelling and potential direction for the future.

This innovative intersection holds significant promise for enhancing key aspects of RGFT, including visualization of internal conflicts, patient engagement, opportunities for safe role-playing, and personalization of therapeutic experiences. By making abstract concepts such as internalized reference groups more tangible and interactive, AI avatars can facilitate deeper self-awareness and promote more effective processing of relational dynamics.

However, responsible and effective implementation of this integration requires careful consideration of the associated ethical challenges and implications, and looking ahead, further research is essential to fully understand the effectiveness, optimal implementation strategies, and long-term impact of AI-assisted RGFT.

Algorithms and Data for Creating Avatars in RGFT

To create a realistic AI avatar of a specific person, various algorithms and data can be used:

Algorithms:

Generative Adversarial Networks (GANs): These are powerful for creating realistic images, including faces, by having two neural networks compete against each other.

Variational Autoencoders (VAEs): Similar to GANs, VAEs can learn the underlying distribution of data and generate new samples, making them suitable for avatar creation.

3D Morphable Face Models (3DMM): These models use Principal Component Analysis (PCA) to create low-level representations of 3D faces, allowing the creation of avatars with specific facial features.

Neural Networks (including Deep Learning models): These are widely used to analyze facial features, expressions, and movements from input data to create realistic avatars.

Text-to-Avatar Algorithms:

These algorithms use natural language processing (NLP) to interpret textual descriptions of a person and create a corresponding visual avatar.

Image-to-Avatar Algorithms:

These algorithms analyze uploaded images (photos or videos) to extract facial features and create a digital representation. Techniques such as facial recognition are used here.

Motion Capture:

For more dynamic avatars, algorithms analyze motion capture data to replicate realistic body movements and gestures.

Data:

Images (Photos): High-quality images of the specific person from different angles (front, side profiles, etc.) and with different facial expressions are essential for training the AI model to accurately represent their appearance. More images generally lead to better results.

Video clips: Short video recordings of a person can be used to capture their mannerisms, expressions, and the way they move, leading to more realistic avatars.

Audio recordings: To create a talking avatar, audio samples of the person’s voice are needed. This allows the avatar to speak with a similar tone, pitch, and accent. Advanced techniques can even replicate a voice from a short audio sample.

Text descriptions: Detailed textual descriptions of a person's physical features, personality, and communication style can be used as input to create text for an avatar.

Facial feature data: Some advanced methods may use specific data points related to facial geometry and texture to create highly accurate avatars.

The specific algorithm and data requirements will depend on the level of realism desired, the intended use of the avatar, and the capabilities of the chosen AI avatar creation platform or software.

Many platforms offer user-friendly interfaces that guide the process of uploading data and customizing the avatar.

To reflect a person’s inner attitudes and characteristics for creating an AI avatar, focus on collecting data that captures their personality, communication style, and emotional nuances.

Here are some ways to approach this:

Behavioral data through interviews and questionnaires:

Structured interviews: Conduct interviews with the person, focusing on their typical responses to different situations, their opinions on different issues, their values, and their general outlook on life. Open-ended questions should be used to encourage detailed responses.

Personality questionnaires: Use assessments of established personality questionnaires that the person can complete. This can provide a framework for understanding their core personality traits.

Scenario-based questions: Present hypothetical scenarios and ask how the person is likely to respond, both emotionally and behaviorally. This can reveal their attitudes and decision-making processes.

Communication style data:

Conversational recordings (with consent): Record the person in natural conversations with friends or family members (with their explicit consent). Analyze these recordings for their typical tone of voice, rate of speech, use of specific phrases or slang, and overall communication patterns.

Written Communication Analysis: Analyze their emails, messages, or social media posts (again, with consent and respecting privacy) to understand their writing style, vocabulary, and typical emotion expressed.

Linguistic Analysis: Use natural language processing (NLP) techniques to identify patterns in their language that may reflect their inner characteristics, such as sentiment, interests, and communication style.

Emotional expression data:

Video recordings (with consent): Record the person in different emotional states (e.g., happy, sad, angry, surprised). Analyze their facial expressions, body language, and tone of voice in these states.

Self-reported emotions: Ask the person to describe how they typically feel in different situations and how they express these feelings.

Emotional language analysis: Through their written and spoken communication, these are words and phrases that indicate their emotional state and typical emotional responses.

Preferences and Values:

Direct inquiry: Ask about their likes, dislikes, interests, hobbies, and what they value in life and relationships.

Choice analysis: Observe or ask about their preferences in various areas, such as music, movies, books, and activities. These choices can often reflect underlying attitudes and values.

Feedback from others (consensus):

Interviews with significant others: With the person’s permission, interview close friends or family members to gather their perspectives on the person’s personality, attitudes, and typical behaviors.

Ethical considerations:

Informed consent: It is essential to obtain explicit and informed consent from the person for the collection and use of their personal data to create the avatar. Clearly explain how the data will be used and who will have access to it.

Privacy and Security: Ensure that all collected data is stored securely and in accordance with privacy regulations. Anonymize data where possible.

Transparency: Be transparent with the individual about the capabilities and limitations of the AI avatar and the process of creating it.

Avoid bias: Be aware of potential biases in the data collection process and in the AI algorithms used to create the avatar. Strive for a balanced and accurate representation.

By collecting a wide range of data points in these categories, you can build a more comprehensive profile of a person’s internal attitudes and characteristics, leading to a more nuanced and effective AI avatar for use in RGFT. Remember that capturing the full complexity of a human personality is a challenging task, and the avatar will likely be a partial representation rather than a perfect replica.

Below is a detailed workflow for creating an optimal AI avatar of a concrete person for use in Reference Group Focused Therapy (RGFT):

Step 1: Initial Consultation and Goal Setting

Therapist-Patient Discussion:

Goal: Clearly define the purpose of the avatar and the specific attitudes and internal characteristics to be reflected.

Questions for the client:

"Who is the specific person you would like this avatar to represent?"

"What is your relationship with this person?"

“In what specific ways do you feel their attitudes or internal characteristics affect you?”

“What are this person’s core attitudes or beliefs that you want the avatar to reflect?” (e.g., skeptical, supportive, critical, optimistic)

“What are this person’s core personality traits?” (e.g., introverted, extroverted, anxious, calm)

“Are there any specific emotional tendencies or patterns that this person has?” (e.g., tendency to worry, easily express excitement, easily get frustrated)

“How do you envision interacting with this avatar in our therapy sessions?”

“Are there any specific scenarios or interactions you would like to explore with this avatar?”

“What are your expectations of this avatar?”

“Are there any aspects of this person’s attitudes or internal characteristics that you would not want the avatar to reflect?”

Guidelines for the therapist:

Clearly explain the purpose of using AI avatars in RGFT and how it can help with understanding internalized reference groups.

Emphasize that the goal is to create a useful therapeutic tool, not a perfect replica.

Discuss the ethical considerations involved, especially regarding privacy and potential biases.

Step 2: Collect Data – Behavioral Data

Structured Interviews with the Client:

Goal: To collect specific examples of the person’s behavior in different situations.

Detailed Questions for the Client:

"Can you describe a specific situation where this person reacted in a typical way? What happened, what did they do, and what did they say?"

"How does this person typically handle stressful situations?"

"How does he or she typically respond to good news?"

"What are his or her typical reactions to disagreements or conflicts?"

"Can you think of a time when this person's core beliefs were evident in his or her actions?"

“What are some of their habits or routines that might reflect their internal state?”

“How does this person interact with people they know well versus strangers?”

“Are there specific triggers that tend to elicit certain behaviors or attitudes from this person?”

Instructions for the therapist:

Encourage the client to provide vivid, detailed descriptions.

Ask for multiple examples in different contexts.

Take detailed notes, focusing on the behaviors observed and using direct quotes whenever possible.

Step 3: Data Collection – Communication Style Data

Analysis of Verbal Communication (Client's Perspective):

Goal: To understand the person's typical speech pattern.

Detailed Questions for the Client:

"How would you describe this person's tone of voice? (e.g., warm, sarcastic, direct, gentle)"

"What is their typical speech rate? (e.g., fast, slow, deliberate)"

"Does he use specific phrases, idioms, or slang often?"

"Is his language formal or informal?"

"Does he tend to ask a lot of questions or make more statements?"

“How does he usually start and end conversations?”

“Does he use humor? If so, what kind?”

“Are there certain topics he tends to talk about often?”

Instructions for the therapist:

Notice the nuances in the client’s descriptions.

Ask for examples of typical phrases or expressions.

Analysis of written communication (if applicable and agreed):

Goal: To understand the person’s writing style.

Client Instructions (if applicable and with consent):

"If you have access to this person's emails or messages, would you be willing to share some examples (while respecting their privacy) that might illustrate their writing style?"

Therapist Instructions:

Analyze the written examples provided for vocabulary, sentence structure, tone, and sentiment.

Note recurring themes or topics.

Step 4: Data Collection – Emotional Expression Data

Gathering information about emotional expression (client's perspective):

Goal: To understand how the person typically expresses their emotions.

Detailed questions for the client:

"How does this person typically show happiness?"

"How does he/she express sadness or disappointment?"

"What does his/her anger look like?"

"How does he/she show surprise or excitement?"

"Does he/she typically express his/her emotions, or is he/she more reserved?"

“Can you describe his typical facial expressions when he feels a particular emotion?”

“What body language does he use to convey his feelings?”

“How does his tone of voice change when he feels different emotions?”

Instructions for the therapist:

Encourage the client to recall specific instances.

Notice any differences between verbal and nonverbal expressions.

Step 5: Collecting Data – Preferences and Values

Exploring Preferences and Values with the Client:

Goal: To understand the person’s core beliefs and what is important to them.

Detailed questions for the client:

“What are this person’s hobbies or interests?”

“What kind of music, movies, or books does he enjoy?”

“What do he value most in life? (e.g., family, career, honesty, kindness)”

“What are his views on important social or ethical issues (if relevant to the goals of the therapy)?”

“What are the things he really likes or dislikes?”

“What are their goals and aspirations?”

“What kind of sense of humor does they have?”

Instructions for the therapist:

Listen for clues about the person’s moral compass and guiding principles.

Step 6: Gathering Data – Feedback from Others (Optional and Consented)

Considering Input from Significant Others (With Target’s Consent):

Goal: To gain additional perspectives and validate the client’s observations.

Instructions for the therapist:

Discuss with the client the potential benefits and ethical considerations of gathering feedback from others who know the target well.

If the client agrees and the target agrees, identify people who can provide valuable insights.

Develop interview questions similar to those used with the client, tailored to the relationship the interviewee has with the target.

Sample questions for others:

"How would you describe his personality?"

"What are some of the attitudes or beliefs that define him?"

"How does he typically communicate?"

"What are his typical emotional responses to different situations?"

Step 7: Data Processing and Analysis

Synthesis of Collected Data:

Instructions for the Therapist:

Review all collected data (interview notes, written examples, feedback from others).

Identify recurring patterns, key characteristics, and any inconsistencies or nuances.

Create a comprehensive profile of the target person’s internal attitudes and characteristics.

Prioritize the most relevant aspects based on the therapeutic goals.

Apart of the above, an embodiment of the specific internalized character through a patient while on a hot chair during the RGFT session can serve as a valuable source for the creation of an avatar as it contains the character’s attitudes, nonverbal language, and its dominant emotional style (in a sense as a patient perceives it).

Step 8: Avatar Creation and Customization

Choosing an AI Avatar Platform:

Guidelines for the Therapist:

Choose an AI avatar creation platform that offers the necessary customization options for appearance, voice, and potential behavior.

Consider platforms that allow for fine-tuning of facial features, expressions, and voice.

Avatar Appearance Customization:

Guidelines for the therapist (with client input):

Use collected behavioral data and client descriptions to guide visual customization of the avatar.

Adjust facial features, hairstyle, clothing, and other details to resemble the target person as closely as possible based on the client’s mental representation.

Voice Selection or Duplication:

Guidelines for the therapist (with client input):

If the platform offers voice selection, select a voice that closely matches the tone, pitch, and speaking style of the target person.

Explore options for voice cloning if available and ethically permissible, using audio samples where possible and with consent.

Programming Behavioral and Communication Styles (if platform allows):

Instructions for the therapist (with client input):

If platform allows for programming or scripting of avatar behavior, use data collected on communication style and typical responses to guide these settings.

Focus on incorporating key phrases, typical responses to specific prompts, and general communication patterns.

Incorporating emotional expressions (if platform allows):

Therapist guidelines (with client input):

Use platform features to allow the avatar to display a range of emotions through facial expressions and body language typical of the target individual.

Step 9: Test and refine

Initial client test:

Therapist guidelines:

Present the avatar to the patient in a therapy session.

Engage in interactions with the avatar, eliciting responses based on the data collected and the therapeutic goals.

Questions to the client:

"How well does this avatar capture the look of…..?"

"Does the voice sound familiar?"

"How realistic are the avatar's reactions and responses?"

"Does the avatar's communication style feel?"

"Are there aspects of the avatar that feel inaccurate or need adjustment?"

"How comfortable do you feel interacting with this avatar?"

Refinement:

Therapist Instructions:

Based on client feedback, make any necessary adjustments to the avatar’s appearance, voice, behavior, and communication style.

Repeat the testing process until the client feels the avatar is a reasonably accurate and helpful representation.

Step 10: Ethical Review and Consent

Final Review of Ethical Considerations:

Therapist Instructions:

Reiterate the limitations of the AI avatar and ensure the client understands that it is a tool for exploration, not a perfect simulation or replacement of the actual person.

Confirm that the client has given informed consent to the creation and use of an avatar for therapy.

Discuss data privacy and security measures related to the avatar and all data collected.

By following this detailed workflow, therapists can systematically collect the necessary data and use AI avatar creation platforms to create highly personalized and effective tools for peer-centered therapy.

Remember that ongoing communication and feedback from the client are essential throughout the process.

That's it for now,

Yours,

Dr. Igor Salganik and Prof. Joseph Levine