top of page
Search

Artificial Intelligence within the Mental Health Domain

Pooja Adur

An investigation of the benefits, risks, and implications of using intelligent user interfaces within the mental health domain to determine whether its uses should be utilized and to what extent.


Figure 1: Ellie, the face of SimSensei, using audiovisual sensors to detect nonverbal cues like low gaze attention or high body activity (like rapid hand gestures). She can then use that information to alter her follow up questions appropriately.



Abstract

Mental health, specifically anxiety and depression, is a rapidly growing problem in the Unites States. Similarly, there seems to be a heightened interest in the creation of intelligent user interfaces which can be used within this domain. There are numerous products being created, some of which assist live psychotherapists, while some are being built to autonomously work with patients. The problem is that there has not been significant consideration of both the benefits and challenges within a single paper on this topic. This paper focuses on the benefits, challenges, and implications to ascertain whether the utilization of IUIs should be employed and to what extent within the mental health domain.

Mental Health: Anxiety and Depression

There are many types of anxiety disorders; however, the most common are generalized, social, panic, and phobia anxiety disorder (Anxiety Disorders, 2017). Currently, 18% of adults and 8% of children and teenagers have anxiety disorders, while 7% of adults suffer from depressive episodes (Anxiety Disorders, 2017; Depression, 2017). These percentages may seem small; however, they have a vastly negative impact on the quality of life for these individuals. People suffering from anxiety or depression are negativity impacted both physiologically and socially. They are more suspectable to viral infections because anxiety and depression can hinder the strength of their immune system and slow down their bodies healing capabilities (Picard & Healey, 1997 ). Their quality of life also deteriorates regarding work, spending habits, social behavior, self-esteem, leisure activities, and friends and family (Barrera & Norton, 2009). Depression can also have a large negative impact on an individual’s eating and sleeping behavior, as well their education, career, and relationships. It can also negatively impact an individual’s motivation and ability to concentrate, resulting in a decrease in their productivity. Anxiety and depression are two different chronic diseases both of which individually have a negative impact on peoples’ lives; however, they also frequently coexist in a single person as 60% of individuals who have anxiety will also have depression and vice versa (Salcedo, 2018). Comorbidity is the term that defines the existence of two chronic diseases in a single patient, and such a combination has an even larger negative impact on an individual’s quality of life, which is why seeking professional help to receive coping strategies and treatment for managing these disorders is crucial (Barrera & Norton, 2009).


Although it is vital to seek professional help, many individuals suffering from anxiety and/or depression abstain from doing so for reasons such as cost, access, and stigma. Seeing a psychotherapist can be extremely costly, and some individuals face a location-based constraint as there may not be a mental health professional available to them within a reasonable distance. Additionally, some cultures do not acknowledge the existence of mental health issues, while others may view it as a sign of weakness. These are factors that can potentially be addressed with approaches utilizing intelligent user interfaces (IUIs).


Current Use and Benefits in Treating Mental Health Issues

Intelligent user interfaces (IUIs) have already been utilized within the medical field with expert system, like MYCIN, which was created for clinical decision making (Cawsey, 1998). As technology continues to increase, IUIs with new capabilities are slowly making their way into the mental health field. Two such examples of current technologies include SimCoach and SimSensei (Luxton, 2014).


Current Uses

One of these IUI technologies is SimCoach, a web based virtual human agent (VHA) which was designed to mimic human gestures, speech, and emotion to make patients more comfortable and elicit information from them (Rizzo, et al., 2016, pp. 59-64). The creation of SimCoach was inspired by military service members and their families who were suffering from anxiety and depression yet were not seeking professional care (Rizzo, et al., 2016). Seeking help was viewed as a form of weakness and was therefore stigmatized (Rizzo, et al., 2016). Additionally, the lack of availability and accessibility were also barriers that prevented individuals from getting the help that they needed (Rizzo, et al., 2016). To make individuals feel more comfortable, they are given the opportunity to choose the gender of the virtual human that they would speak with, and then they can anonymously speak about their past health history and current mental health concerns so that they can be provided with resources for further care. Originally SimCoach did not provide treatment directly to patients. Rather, its goal was to break down the patient’s barrier about visiting a live psychotherapist and provide those resources to them. The future goal for SimCoach is for patients to be able to repeatedly seek assistance, coping mechanisms, and treatment from it; however, this aspect is still being developed and has a couple of implications which are further discussed later in this paper.


SimSensei (Figure 1) is another virtual human agent (VHA) which provides face to face interactions for clinical interviewing to assess and detect when a patient is showing signs of psychological distress. It allows individuals who feel stigmatized or may not have a psychotherapist within a reasonable distance to seek help and get a proper assessment without having to see a live psychotherapist. Ellie, the face of SimSensei, uses natural language to communicate and real-time audiovisual sensors, to detect a patient’s body gestures, facial expressions, and speech patterns and undulations to identify nonverbal behaviors that indicate psychological distress. She is able to listen to what her patients say and simultaneously analyze the data collected from the audiovisual sensors to ask appropriate follow up questions, resulting in increased rapport which can assist in detecting mental health issues (Rizzo, et al., 2016, pp. 64-70).


BioBeats is a new IUI that is concentrating on quantifying mental health by collecting psychological, neurological, and physiological data from “cardiovascular to movement, brain function to several validated measures of depression and anxiety, and others” (Sherriff, 2019). It measures affective patterns through the existing sensors in smartphones and an individual’s wearable devices to detect stress patterns (Sherriff, 2019). BioBeat’s main goal is to show individuals their distinctive stressors, how it can impact their body, mind, and life, and ways in which they can cope (Sherriff, 2019). This is a relatively new piece of technology, so it is not available to the general population as it currently lacks the knowledge and infrastructure to refer individuals to qualified mental health professionals if they have severe anxiety, depression, or comorbidity (Weiner, 2018).

Benefits of the Utilization of Intelligent User Interfaces

SimCoach, SimSensei, and BioBeats are three examples of many different IUIs that are being used today. These pieces of technology have their share of benefits and challenges, and in the following section the benefits will be explored in more detail.


As noted earlier, stigma is one of the big reasons that people refrain from seeking professional help even if they are suffering from anxiety and/or depression. Stigma is a negative stereotype that is associated with an event or situation. People often fear stigmatization when they must go and interact with a live person to discuss their mental health problems. To combat that fear of judgement, IUIs like SimCoach can be utilized to simulate ‘person to person’ interactions. This technology can be very beneficial for patients and can provide them with the necessary assessments, recommendations, and referrals whenever it is most convenient for them. Individuals, who do not seek professional help because of the stigma, may feel more comfortable talking to an IUI as it affords them the privacy they need to receive care for their anxiety and/or depression (Luxton, 2014). Virtual human agents like SimCoach and SimSensei can help numerous people seek care without having to go to a ‘live’ psychotherapist and give them the opportunity to take the first steps needed to improve their mental health and quality of life.


Even when an individual wholeheartedly wants to seek professional help, there can still be barriers like cost and location that can prevent them from doing so. Some psychotherapists do not accept insurance and even when they do, co-pays alone may be extremely costly. Some patients also face a location-based constraint as there may not be any mental health professionals that are located within a reasonable distance, preventing them from seeking the help they need. Consistency is key when it comes to receiving proper treatment for anxiety and/or depression and people who are located a considerable distance away can be less motivated or able to regularly see their psychotherapist. With virtual human agents, individuals can access their therapy directly from their computer, which can encourage consistent therapy sessions resulting in an increased probably of a successful treatment. The creation of IUIs for mental health assistance can reach individuals who may not be able to afford or physical attend traditional therapy.


There is currently a large shortage of psychotherapists, resulting in waiting periods of 1-4 weeks for individuals who are seeking professional medical help (Dampier, 2018). As the number of individuals suffering from anxiety and/or depression increases, additional cognitive strain is being placed on psychotherapists to meet that increased demand, resulting in burnout (Weiner, 2018). This overuse of their cognitive resources can result in fatigue, inattentive behavior, and increased human error (Chen, Castro-Alonso, Paas, & Sweller, 2018). Intelligent user interfaces do not experience the burnout that humans do; therefore, utilizing IUIs can decrease human error while enhancing the capabilities of mental health professionals. These technologies have capabilities beyond those of human health professionals and can detect subtle changes in patients that may not noticeable to a human observer. Although mental health professionals understand the clinical relevance of the increase in internal body temperature of thier patient, they have no way of detecting that increase unless the patient informs them, or they run a test. Intelligent user interfaces that have capabilities like infrared imaging to detect body temperature, sensors that measure heart rate, and optical sensing to detect subtle facial expressions or eye blinking, can immensely assist psychotherapists by compensating for their human limitations. The IUI can assist by contributing additional information that would otherwise not have been collected during a patient’s session, allowing the psychotherapist to further understand their patient. The technology can also be utilized to conduct sessions with patients autonomously, extending therapy beyond scheduled sessions (Luxton, 2014).


Virtual and augmented reality are both currently being used to combat different challenges within mental health. Virtual reality (VR) can completely immerse patients in a computer-generated therapist’s office, where they can interact with a virtual human psychotherapist to attend their therapy sessions (Luxton, 2016). It gives individuals the opportunity to receive real-time therapy through a virtual therapist when they are either not able or do not want to visit a mental health professional. Augmented reality (AR) or ‘mixed reality’ is currently being utilized to help expose patients to the stimuli that trigger their anxiety to unpair the stimuli from those feelings. This is a method called Prolonged Exposure Therapy, and AR allows patients to participate in it within a safe and controlled environment (Luxton, 2016).


In an earlier section of this paper, examples were given of IUIs that were able to understand human emotion and then act accordingly. This is achieved through affective computing, which is a sub-discipline within computer science that concentrates on giving IUIs the capability to detect, categorize, and respond to human emotion (Luxton, 2016). This technology can be found in wearables or in ambient sensors and can be very beneficial for numerous reasons. It can be used to detect an individual’s affective patterns, such as facial expressions, movement, gestures, heart rate, sleep patterns, voice, skin conductivity, and social interactions to diagnose anxiety or depression in unknowing individuals (Luxton, 2016, p. 7). An example of this is SimSensei who uses affective computing to determine indicators of psychological distress in individuals during assessment interviews (Stratou, et al., 2015 ). BioBeats, although relatively new, is also great use of affective computing which can assist individuals in detecting their stressor and indicators of anxiety and/or depression. Another beneficial use for affective wearables is that they gather data about patients as they go about their normal daily activities. This allows the patient and their psychotherapist to monitor their affective patterns outside of therapy to gain more insights about their stressors, which can help in monitoring and dynamically personalizing treatment for each patient.


Currently most psychotherapists create a tailored treatment and monitoring plan for their patients based on their specific experiences and needs without any IUI assistance. That personalization can be made easier by connecting multiple sources of information to get a dynamic understanding of the patient. One of the many benefits of an IUI is its ability to access numerous sources of information; however, to do so there needs to be a database large enough to store all of it. Wireless and cloud technologies give us the opportunity to have intelligent agents leverage numerous sources of information to complete their task. In the case of anxiety and depression, intelligent agents can leverage their expert knowledge in psychology and psychiatry and access their patient’s health records, the data on their wearables, and even affective patterns from atmospheric sensors to better understand their patient’s condition (McShane, Beale, Nirenburg, Jarrell, & Fantry, 2012). By gathering that data, an IUI can effectively and actively assess, diagnose, treat, and monitor its patients. During and after receiving treatment, continuing to monitor and work towards improving your mental health can be difficult. IUIs in wearables, as noted earlier, can help by monitoring and detecting physiological changes in patients and assist by reminding them of techniques they can use alleviate their symptoms (Luxton, 2016). Psychotherapists can use this technology in many ways to personalize treatment and provide a way to always make a patient feel supported.


Intelligent user interfaces can reduce the workload on psychotherapists by having virtual human agents, like SimSensei, conduct the clinical assessment interviews before passing the information on to the psychotherapist. This gives them the ability to perform their job much more effectively and potentially help more patients, considering the cognitive load required to work with each patient may be lowered due to the assistance. IUIs are also breaking down the barriers that some people have about psychotherapists when they are suffering from anxiety and/or depression and allows them to take the first step to improving their mental health


Challenges and Implications of IUIs in Mental Health Care

Numerous benefits were discussed in the previous section regarding the utilization of IUIs within the mental health domain. However, there are some challenges with serious implications that can affect the diagnosis and treatment of patients which need to be considered.


A considerable challenge that IUIs, like virtual human agents, face is the ability to develop empathy and a human connection that would form between a live psychotherapist and their patient. Empathy is the ability to understand, share, and respond to the feelings of others in a supportive manner, and is how psychotherapists can understand their patients’ emotions rationally and offer support. The way that people covey empathy and form human relationships are through nonverbal cues, like gaze and expressions, which take years to cultivate (Frischen, Bayliss, & Tipper, 2007). The development of gaze perception begins in infancy and is a large factor in an individual’s social cognitive development (Frischen, Bayliss, & Tipper, 2007). This means that for a VHA to be successful, it needs to recognize and understand the implications of the emotions that their patients are feeling, and then appropriately react in a visual and verbal manner that is recognizable to their patient. VHAs, like SimSensei, are already capable of emotion recognition using affective computing and multimodal audiovisual sensors, and although the detectable emotions are currently relatively basic, they are improving (D'Mello, 2016).


The expression of emotion is a lot more difficult to master and extremely important in order to encourage trust, realism, and engagement. There are two factors in the expression of emotion: the ability to display an emotion in the appropriate manner and the capability to display an emotion in an applicable context (D'Mello, 2016). The VHA needs to be able to synchronize every facial expression, hand movement, and eye movement to correctly convey an emotion in a specific context, which can be very difficult (D'Mello, 2016). What makes it more challenging is that certain affective patterns differ based on the patient’s culture, expectations, and upbringing. There are many cultural implications that can factor into each patient’s experience, which can impact their assessment, diagnosis, and treatment. This is something that a professional psychotherapist must take time to learn and understand. Not understanding a patient’s background may potentially cause an inaccurate diagnosis resulting in erroneous treatment recommendations. This is information that can be hard to input into an intelligent user interface. It is stated that IUIs are now capable of adapting their mannerisms, such as eye contact and speech dialect, based on a patient’s culture (Luxton, 2016). The problem here is that culture is not textbook. It can vary based on nationality, localities, social class, gender, education, and profession, to name a few, especially in the United States, which is the melting pot of the world. A first generation American can grow up with the influences of their ethnic culture which can then be blended with their experiences growing up in a certain area of the United States. These cultural nuances vary based on the individual, which makes it difficult to predict, understand, and input into a VHA’s database.


Although VHAs use affective computing to recognize and mimic human emotion in the right context, the experience and wisdom that a live psychotherapist contributes is something that a VHA cannot have, which can be an integral part of creating a therapeutic bond and rapport. Additionally, there are a colossal number of nonverbal cues, that vastly differ based on culture, within human behavior and communication that a VHA would need knowledge off to accurately recognize and appropriately react to a patient’s emotions and actions (D'Mello, 2016). The difficulty here is creating a knowledge base full of the numerous different robust descriptors of human behavior which is influenced by context and can be inputted imprecisely (Luxton, 2014).This database is still growing and requires a lot of time, effort, and money to capture the data that is necessary for the VHA to function (D'Mello, 2016). Additionally, it creates the possibility that a VHA may incorrectly respond to a patient if they do not have the data required to properly infer the meaning of an affective pattern. An error may be small but can potentially have negative implications on not only the patient’s experience during therapy but also on their assessment and treatment plan if a VHA was conducting this session autonomously. That error can lead to issues regarding liability. Who is responsible if a VHA makes an error which leads to a negative side effect in a patient’s life? It is incorrectly assumed that intelligent user interfaces are not prone to error because they are not plagued by the same cognitive restrictions as humans. However, they are modeled by humans who are ‘error prone’ and their knowledge base is still growing, so errors are possible. This is another concern that should be considered when discussing the extent to which an IUI should be responsible for someone’s mental health.


The benefits of connecting numerous intelligent agents together are great; however, by connection all that information it also makes it more vulnerable. The more information that is connected and collected, the more privacy becomes an issue that needs to be discussed. If a VHA is working with patients autonomously, rather than assisting a psychotherapist, they would need to store notes and personal information discussed during each session in their database for subsequent appointments to be meaningful and efficient. However, this poses several problems with anonymity and privacy. To store and categorize a patient’s information, an account would need to be created (Bennet & Doub, 2016). This eliminates the anonymity which attracted certain patients to consult with virtual human agents in the first place. This also causes a problem in terms of privacy because that database, if it were to be hacked, would expose hundreds of patient’s private session notes and past medical history all at once.


Conclusion

As discussed in this paper, intelligent user interfaces, like SimCoach, SimSensei, and BioBeats, can provide excellent benefits for numerous issues which are plaguing the industry, like stigma, location, cost, and burnout to name a few. However, people should be aware of the challenges as well. For the purposes of this paper, trust was not discussed in detail as one of the challenges that can be faced; however, this would be a great topic to delve deeper into for the future. Based on the benefits and challenges uncovered during the literature review, it is recommended that IUIs should be employed within the mental health industry to assist live psychotherapists; however, they should not be responsible for the mental health of an individual completely. Handing over all the responsibility to an intelligent agent can be dangerous considering the challenges discussed. Perhaps with new technology that may change; however, currently there are issues surrounding emotion, culture, privacy, and liability which cannot be ignored.


Intelligent User Interfaces for Anxiety and Depression: Benefits and Challenges



Abstract

This paper focuses on how intelligent user interfaces (IUIs) are employed within the mental health domain and the potential costs and benefits of utilizing them. Mental health issues, like anxiety and depression, are a growing problem that many people face in the United States and around the world. Although it is a rapidly growing problem, people still refrain from receiving professional help for reasons such as cost, access, and stigma. Seeing a therapist or psychiatrist can be extremely costly, and some individuals face a location-based constraint as there may not be a mental health professional available to them within a reasonable distance. Additionally, some cultures do not acknowledge the existence of mental health issues, while others may view it as a sign of weakness. A literature review was conducted looking at both currently utilized IUI products as well as research within the mental health domain to ascertain the positive and negative implications of using IUIs. This paper investigates the risks, benefits, and implications to determine whether the use of IUIs should be utilized and to what extent within the mental health domain.


Introduction

Mental health, specifically anxiety and depression, is a rapidly growing problem in the Unites States. Similarly, there seems to be a heightened interest in the creation of intelligent user interfaces which can be used within this domain. There are numerous products being created, some of which assist live psychotherapists, while some are being built to autonomously work with patients. The problem is that there has not been significant consideration of both the benefits and challenges within a single paper on this topic. This paper focuses on the benefits, challenges, and implications to ascertain whether the utilization of IUIs should be employed and to what extent within the mental health domain.

Mental Health: Anxiety and Depression

There are many types of anxiety disorders; however, the most common are generalized, social, panic, and phobia anxiety disorder (Anxiety Disorders, 2017). Currently, 18% of adults and 8% of children and teenagers have anxiety disorders, while 7% of adults suffer from depressive episodes (Anxiety Disorders, 2017; Depression, 2017). These percentages may seem small; however, they have a vastly negative impact on the quality of life for these individuals. People suffering from anxiety or depression are negativity impacted both physiologically and socially. They are more suspectable to viral infections because anxiety and depression can hinder the strength of their immune system and slow down their bodies healing capabilities (Picard & Healey, 1997 ). Their quality of life also deteriorates regarding work, spending habits, social behavior, self-esteem, leisure activities, and friends and family (Barrera & Norton, 2009). Depression can also have a large negative impact on an individual’s eating and sleeping behavior, as well their education, career, and relationships. It can also negatively impact an individual’s motivation and ability to concentrate, resulting in a decrease in their productivity. Anxiety and depression are two different chronic diseases both of which individually have a negative impact on peoples’ lives; however, they also frequently coexist in a single person as 60% of individuals who have anxiety will also have depression and vice versa (Salcedo, 2018). Comorbidity is the term that defines the existence of two chronic diseases in a single patient, and such a combination has an even larger negative impact on an individual’s quality of life, which is why seeking professional help to receive coping strategies and treatment for managing these disorders is crucial (Barrera & Norton, 2009).

Although it is vital to seek professional help, many individuals suffering from anxiety and/or depression abstain from doing so for reasons such as cost, access, and stigma. Seeing a psychotherapist can be extremely costly, and some individuals face a location-based constraint as there may not be a mental health professional available to them within a reasonable distance. Additionally, some cultures do not acknowledge the existence of mental health issues, while others may view it as a sign of weakness. These are factors that can potentially be addressed with approaches utilizing intelligent user interfaces (IUIs).

Current Use and Benefits in Treating Mental Health Issues

Intelligent user interfaces (IUIs) have already been utilized within the medical field with expert system, like MYCIN, which was created for clinical decision making (Cawsey, 1998). As technology continues to increase, IUIs with new capabilities are slowly making their way into the mental health field. Two such examples of current technologies include SimCoach and SimSensei (Luxton, 2014).

Current Uses

One of these IUI technologies is SimCoach, a web based virtual human agent (VHA) which was designed to mimic human gestures, speech, and emotion to make patients more comfortable and elicit information from them (Rizzo, et al., 2016, pp. 59-64). The creation of SimCoach was inspired by military service members and their families who were suffering from anxiety and depression yet were not seeking professional care (Rizzo, et al., 2016). Seeking help was viewed as a form of weakness and was therefore stigmatized (Rizzo, et al., 2016). Additionally, the lack of availability and accessibility were also barriers that prevented individuals from getting the help that they needed (Rizzo, et al., 2016). To make individuals feel more comfortable, they are given the opportunity to choose the gender of the virtual human that they would speak with, and then they can anonymously speak about their past health history and current mental health concerns so that they can be provided with resources for further care. Originally SimCoach did not provide treatment directly to patients. Rather, its goal was to break down the patient’s barrier about visiting a live psychotherapist and provide those resources to them. The future goal for SimCoach is for patients to be able to repeatedly seek assistance, coping mechanisms, and treatment from it; however, this aspect is still being developed and has a couple of implications which are further discussed later in this paper.

SimSensei (Figure 1) is another virtual human agent (VHA) which provides face to face interactions for clinical interviewing to assess and detect when a patient is showing signs of psychological distress. It allows individuals who feel stigmatized or may not have a psychotherapist within a reasonable distance to seek help and get a proper assessment without having to see a live psychotherapist. Ellie, the face of SimSensei, uses natural language to communicate and real-time audiovisual sensors, to detect a patient’s body gestures, facial expressions, and speech patterns and undulations to identify nonverbal behaviors that indicate psychological distress. She is able to listen to what her patients say and simultaneously analyze the data collected from the audiovisual sensors to ask appropriate follow up questions, resulting in increased rapport which can assist in detecting mental health issues (Rizzo, et al., 2016, pp. 64-70).

BioBeats is a new IUI that is concentrating on quantifying mental health by collecting psychological, neurological, and physiological data from “cardiovascular to movement, brain function to several validated measures of depression and anxiety, and others” (Sherriff, 2019). It measures affective patterns through the existing sensors in smartphones and an individual’s wearable devices to detect stress patterns (Sherriff, 2019). BioBeat’s main goal is to show individuals their distinctive stressors, how it can impact their body, mind, and life, and ways in which they can cope (Sherriff, 2019). This is a relatively new piece of technology, so it is not available to the general population as it currently lacks the knowledge and infrastructure to refer individuals to qualified mental health professionals if they have severe anxiety, depression, or comorbidity (Weiner, 2018).

Benefits of the Utilization of Intelligent User Interfaces

SimCoach, SimSensei, and BioBeats are three examples of many different IUIs that are being used today. These pieces of technology have their share of benefits and challenges, and in the following section the benefits will be explored in more detail.

As noted earlier, stigma is one of the big reasons that people refrain from seeking professional help even if they are suffering from anxiety and/or depression. Stigma is a negative stereotype that is associated with an event or situation. People often fear stigmatization when they must go and interact with a live person to discuss their mental health problems. To combat that fear of judgement, IUIs like SimCoach can be utilized to simulate ‘person to person’ interactions. This technology can be very beneficial for patients and can provide them with the necessary assessments, recommendations, and referrals whenever it is most convenient for them. Individuals, who do not seek professional help because of the stigma, may feel more comfortable talking to an IUI as it affords them the privacy they need to receive care for their anxiety and/or depression (Luxton, 2014). Virtual human agents like SimCoach and SimSensei can help numerous people seek care without having to go to a ‘live’ psychotherapist and give them the opportunity to take the first steps needed to improve their mental health and quality of life.

Even when an individual wholeheartedly wants to seek professional help, there can still be barriers like cost and location that can prevent them from doing so. Some psychotherapists do not accept insurance and even when they do, co-pays alone may be extremely costly. Some patients also face a location-based constraint as there may not be any mental health professionals that are located within a reasonable distance, preventing them from seeking the help they need. Consistency is key when it comes to receiving proper treatment for anxiety and/or depression and people who are located a considerable distance away can be less motivated or able to regularly see their psychotherapist. With virtual human agents, individuals can access their therapy directly from their computer, which can encourage consistent therapy sessions resulting in an increased probably of a successful treatment. The creation of IUIs for mental health assistance can reach individuals who may not be able to afford or physical attend traditional therapy.

There is currently a large shortage of psychotherapists, resulting in waiting periods of 1-4 weeks for individuals who are seeking professional medical help (Dampier, 2018). As the number of individuals suffering from anxiety and/or depression increases, additional cognitive strain is being placed on psychotherapists to meet that increased demand, resulting in burnout (Weiner, 2018). This overuse of their cognitive resources can result in fatigue, inattentive behavior, and increased human error (Chen, Castro-Alonso, Paas, & Sweller, 2018). Intelligent user interfaces do not experience the burnout that humans do; therefore, utilizing IUIs can decrease human error while enhancing the capabilities of mental health professionals. These technologies have capabilities beyond those of human health professionals and can detect subtle changes in patients that may not noticeable to a human observer. Although mental health professionals understand the clinical relevance of the increase in internal body temperature of thier patient, they have no way of detecting that increase unless the patient informs them, or they run a test. Intelligent user interfaces that have capabilities like infrared imaging to detect body temperature, sensors that measure heart rate, and optical sensing to detect subtle facial expressions or eye blinking, can immensely assist psychotherapists by compensating for their human limitations. The IUI can assist by contributing additional information that would otherwise not have been collected during a patient’s session, allowing the psychotherapist to further understand their patient. The technology can also be utilized to conduct sessions with patients autonomously, extending therapy beyond scheduled sessions (Luxton, 2014).

Virtual and augmented reality are both currently being used to combat different challenges within mental health. Virtual reality (VR) can completely immerse patients in a computer-generated therapist’s office, where they can interact with a virtual human psychotherapist to attend their therapy sessions (Luxton, 2016). It gives individuals the opportunity to receive real-time therapy through a virtual therapist when they are either not able or do not want to visit a mental health professional. Augmented reality (AR) or ‘mixed reality’ is currently being utilized to help expose patients to the stimuli that trigger their anxiety to unpair the stimuli from those feelings. This is a method called Prolonged Exposure Therapy, and AR allows patients to participate in it within a safe and controlled environment (Luxton, 2016).

In an earlier section of this paper, examples were given of IUIs that were able to understand human emotion and then act accordingly. This is achieved through affective computing, which is a sub-discipline within computer science that concentrates on giving IUIs the capability to detect, categorize, and respond to human emotion (Luxton, 2016). This technology can be found in wearables or in ambient sensors and can be very beneficial for numerous reasons. It can be used to detect an individual’s affective patterns, such as facial expressions, movement, gestures, heart rate, sleep patterns, voice, skin conductivity, and social interactions to diagnose anxiety or depression in unknowing individuals (Luxton, 2016, p. 7). An example of this is SimSensei who uses affective computing to determine indicators of psychological distress in individuals during assessment interviews (Stratou, et al., 2015 ). BioBeats, although relatively new, is also great use of affective computing which can assist individuals in detecting their stressor and indicators of anxiety and/or depression. Another beneficial use for affective wearables is that they gather data about patients as they go about their normal daily activities. This allows the patient and their psychotherapist to monitor their affective patterns outside of therapy to gain more insights about their stressors, which can help in monitoring and dynamically personalizing treatment for each patient.

Currently most psychotherapists create a tailored treatment and monitoring plan for their patients based on their specific experiences and needs without any IUI assistance. That personalization can be made easier by connecting multiple sources of information to get a dynamic understanding of the patient. One of the many benefits of an IUI is its ability to access numerous sources of information; however, to do so there needs to be a database large enough to store all of it. Wireless and cloud technologies give us the opportunity to have intelligent agents leverage numerous sources of information to complete their task. In the case of anxiety and depression, intelligent agents can leverage their expert knowledge in psychology and psychiatry and access their patient’s health records, the data on their wearables, and even affective patterns from atmospheric sensors to better understand their patient’s condition (McShane, Beale, Nirenburg, Jarrell, & Fantry, 2012). By gathering that data, an IUI can effectively and actively assess, diagnose, treat, and monitor its patients. During and after receiving treatment, continuing to monitor and work towards improving your mental health can be difficult. IUIs in wearables, as noted earlier, can help by monitoring and detecting physiological changes in patients and assist by reminding them of techniques they can use alleviate their symptoms (Luxton, 2016). Psychotherapists can use this technology in many ways to personalize treatment and provide a way to always make a patient feel supported.

Intelligent user interfaces can reduce the workload on psychotherapists by having virtual human agents, like SimSensei, conduct the clinical assessment interviews before passing the information on to the psychotherapist. This gives them the ability to perform their job much more effectively and potentially help more patients, considering the cognitive load required to work with each patient may be lowered due to the assistance. IUIs are also breaking down the barriers that some people have about psychotherapists when they are suffering from anxiety and/or depression and allows them to take the first step to improving their mental health.

Challenges and Implications of IUIs in Mental Health Care

Numerous benefits were discussed in the previous section regarding the utilization of IUIs within the mental health domain. However, there are some challenges with serious implications that can affect the diagnosis and treatment of patients which need to be considered.

A considerable challenge that IUIs, like virtual human agents, face is the ability to develop empathy and a human connection that would form between a live psychotherapist and their patient. Empathy is the ability to understand, share, and respond to the feelings of others in a supportive manner, and is how psychotherapists can understand their patients’ emotions rationally and offer support. The way that people covey empathy and form human relationships are through nonverbal cues, like gaze and expressions, which take years to cultivate (Frischen, Bayliss, & Tipper, 2007). The development of gaze perception begins in infancy and is a large factor in an individual’s social cognitive development (Frischen, Bayliss, & Tipper, 2007). This means that for a VHA to be successful, it needs to recognize and understand the implications of the emotions that their patients are feeling, and then appropriately react in a visual and verbal manner that is recognizable to their patient. VHAs, like SimSensei, are already capable of emotion recognition using affective computing and multimodal audiovisual sensors, and although the detectable emotions are currently relatively basic, they are improving (D'Mello, 2016).

The expression of emotion is a lot more difficult to master and extremely important in order to encourage trust, realism, and engagement. There are two factors in the expression of emotion: the ability to display an emotion in the appropriate manner and the capability to display an emotion in an applicable context (D'Mello, 2016). The VHA needs to be able to synchronize every facial expression, hand movement, and eye movement to correctly convey an emotion in a specific context, which can be very difficult (D'Mello, 2016). What makes it more challenging is that certain affective patterns differ based on the patient’s culture, expectations, and upbringing. There are many cultural implications that can factor into each patient’s experience, which can impact their assessment, diagnosis, and treatment. This is something that a professional psychotherapist must take time to learn and understand. Not understanding a patient’s background may potentially cause an inaccurate diagnosis resulting in erroneous treatment recommendations. This is information that can be hard to input into an intelligent user interface. It is stated that IUIs are now capable of adapting their mannerisms, such as eye contact and speech dialect, based on a patient’s culture (Luxton, 2016). The problem here is that culture is not textbook. It can vary based on nationality, localities, social class, gender, education, and profession, to name a few, especially in the United States, which is the melting pot of the world. A first generation American can grow up with the influences of their ethnic culture which can then be blended with their experiences growing up in a certain area of the United States. These cultural nuances vary based on the individual, which makes it difficult to predict, understand, and input into a VHA’s database.

Although VHAs use affective computing to recognize and mimic human emotion in the right context, the experience and wisdom that a live psychotherapist contributes is something that a VHA cannot have, which can be an integral part of creating a therapeutic bond and rapport. Additionally, there are a colossal number of nonverbal cues, that vastly differ based on culture, within human behavior and communication that a VHA would need knowledge off to accurately recognize and appropriately react to a patient’s emotions and actions (D'Mello, 2016). The difficulty here is creating a knowledge base full of the numerous different robust descriptors of human behavior which is influenced by context and can be inputted imprecisely (Luxton, 2014).This database is still growing and requires a lot of time, effort, and money to capture the data that is necessary for the VHA to function (D'Mello, 2016). Additionally, it creates the possibility that a VHA may incorrectly respond to a patient if they do not have the data required to properly infer the meaning of an affective pattern. An error may be small but can potentially have negative implications on not only the patient’s experience during therapy but also on their assessment and treatment plan if a VHA was conducting this session autonomously. That error can lead to issues regarding liability. Who is responsible if a VHA makes an error which leads to a negative side effect in a patient’s life? It is incorrectly assumed that intelligent user interfaces are not prone to error because they are not plagued by the same cognitive restrictions as humans. However, they are modeled by humans who are ‘error prone’ and their knowledge base is still growing, so errors are possible. This is another concern that should be considered when discussing the extent to which an IUI should be responsible for someone’s mental health.

The benefits of connecting numerous intelligent agents together are great; however, by connection all that information it also makes it more vulnerable. The more information that is connected and collected, the more privacy becomes an issue that needs to be discussed. If a VHA is working with patients autonomously, rather than assisting a psychotherapist, they would need to store notes and personal information discussed during each session in their database for subsequent appointments to be meaningful and efficient. However, this poses several problems with anonymity and privacy. To store and categorize a patient’s information, an account would need to be created (Bennet & Doub, 2016). This eliminates the anonymity which attracted certain patients to consult with virtual human agents in the first place. This also causes a problem in terms of privacy because that database, if it were to be hacked, would expose hundreds of patient’s private session notes and past medical history all at once.


Conclusion

As discussed in this paper, intelligent user interfaces, like SimCoach, SimSensei, and BioBeats, can provide excellent benefits for numerous issues which are plaguing the industry, like stigma, location, cost, and burnout to name a few. However, people should be aware of the challenges as well. For the purposes of this paper, trust was not discussed in detail as one of the challenges that can be faced; however, this would be a great topic to delve deeper into for the future. Based on the benefits and challenges uncovered during the literature review, it is recommended that IUIs should be employed within the mental health industry to assist live psychotherapists; however, they should not be responsible for the mental health of an individual completely. Handing over all the responsibility to an intelligent agent can be dangerous considering the challenges discussed. Perhaps with new technology that may change; however, currently there are issues surrounding emotion, culture, privacy, and liability which cannot be ignored.

Bibliography

Anxiety Disorders. (2017, January). Retrieved from National Alliance on Mental Illness: https://www.nami.org/Learn-More/Mental-Health-Conditions/Anxiety-Disorders


Barrera, T. L., & Norton, P. J. (2009). Quality of Life Impairment in Generalized Anxiety Disorder, Social Phobia, and Panic Disorder. Journal of anxiety disorders, 23(8), 1086–1090. doi:doi:10.1016/j.janxdis.2009.07.011


Bennet, C. C., & Doub, T. W. (2016). Expert System in Mental Health Care: AI Applications in Decision-Making and Consultation. In D. D. Luxton, Artificial Intelligence in Behavioral and Mental Health Care (pp. 27- 52). Elsevier.


Cawsey, A. (1998). The Essence of Artificial Intelligence. Prentice Hall Europe.

Chen, O., Castro-Alonso, J. C., Paas, F., & Sweller, J. (2018). Extending Cognitive Load Theory to Incorporate Working Memory Resource Depletion: Evidence from the Spacing Effect. Education Psychology Review, 483-501.


Dampier, C. (2018, October 25). Mental health care appointments often come with a long wait. 3 ways to cope while help is delayed. Retrieved from Chicago Tribune: https://www.chicagotribune.com/lifestyles/sc-fam-mental-health-wait-times-1030-story.html

Depression. (2017, August). Retrieved from National Alliance on Mental Illness: https://www.nami.org/Learn-More/Mental-Health-Conditions/Depression


D'Mello, S. K. (2016). Automated Mental State Detection for Mental Health Care. (D. D. Luxton, Ed.) Artifical Intelligence in Behavioral and Mental Health Care, 117-136.

Frischen, A., Bayliss, A. P., & Tipper, S. P. (2007). Gaze Cueing of Attention. Psychological Bulletin, 133(4), 694-724. Retrieved from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1950440/


Luxton, D. D. (2014). Artificial Intelligence in Psychological Practice: Current and Future Applications and Implications. Professional Psychology: Research and Practice, 45(5), 332-339. doi:http://dx.doi.org/10.1037/a0034559


Luxton, D. D. (2016). An Introduction to Artificial Intelligence in Behavioral and Mental Health Care. In D. D. Luxton (Ed.), Artificial Intelligence in Behavioral and Mental Health Care (pp. 1-26). Elsevier.


McShane, M., Beale, S., Nirenburg, S., Jarrell, B., & Fantry, G. (2012). Inconsistency as a diagnostic tool in a society of intelligent agents. Artificial Intelligence in Medicine, 137-148.


Picard, R., & Healey, J. (1997 ). Affective Wearables. Proceedings of the First International Symposium on Wearable Computers.


Rizzo, A., Shilling, R., Forbell, E., Scherer, S., Gratch, J., & Morency, L.-P. (2016). Autonomous Virtual Human Agents for Healthcare Information Support and Clinical Interviewing. (D. D. Luxton, Ed.) Artificial Intelligence in Behavioral and Mental Health Care, 53-79.


Salcedo, B. (2018, January 19). The Comorbidity Of Anxiety And Depression. Retrieved from National Alliance on Mental Illness: https://www.nami.org/Blogs/NAMI-Blog/January-2018/The-Comorbidity-of-Anxiety-and-Depression


Sherriff, L. (2019, March 25). How One Company Is Using AI To Address Mental Health. Retrieved from Forbes: https://www.forbes.com/sites/lucysherriff/2019/03/25/how-one-company-is-using-ai-to-address-mental-health/#69af84311319


Stratou, G., Morency, L.-P., DeVault, D., Hartholt, A., Fast, E., Lhommet, M., . . . Rizzo, A. (2015 ). A Demonstration of the Perception System in SimSensei, a Virtual Human Application for Healthcare Interviews. International Conference on Affective Computing and Intelligent Interaction , 787-789.


Weiner, S. (2018, February 13). Addressing the escalating psychiatrist shortage. Retrieved from The Associate of American Medical Colleges News: https://news.aamc.org/patient-care/article/addressing-escalating-psychiatrist-shortage/

 
 
 

Comments


bottom of page