an empathic chatbot is not analternative of a mental health professional or a therapist

Surveys conducted on Woebot users by StanfordUniversity indicate a significant improvement with feelingsof depression and anxiety [20]

The therapy aims to turn the patient’snegative thoughts into positive ones

Lexical speech features: These features are boundwith the vocabulary used by the mental healthpatient. Lexical features need the text extractionfrom the speech to predict the patient’s emotions soit can be used on the recorded audio files.• Acoustic speech features: These features are boundwith the pitch, jitter and tone of the mental healthpatient. Acoustic features need the audio data forunderstanding the emotions in the patient’sconversation so it can be used for voice calls withthe patient. The acoustic model will be trained toextract the spectral features from speech signals

he data is gathered from various physical featuressuch as body movements, facial expressions, eye contact andother physical, biological signals

Chatbots can identify emergencies such as suicidalthoughts and escalate or redirect them to appropriateprofessionals

emojis/emoticons

Sentiment Analysis

Natural Language Processing (NLP). NLPdepicts how chatbots translate and understand the patient’slanguage. Using NLP, chatbots can make sense of the spokenor written text and accomplish the tasks like keywordextraction, translation, and topic classification. NLPprocesses the content expressed in natural human languagewith the help of the techniques such as sentiment analysis,facial recognition and voice recognition [10].

Understand and manage the patient’s psychologicalstate and connect them with a health professionalduring unfavourable events.2. 24/7 Instant chat support3. Smart with reactive behaviour such as promptanswering of a question and engage the patients withillness prevention and care tips.4. Easy to install, configure and maintain and iscompatible with various operating systems such asAndroid, iOS and Linux.5. For sensitive health care issues, patients might feelless shame and feel more private.6. Security of personal data is enhanced using differentauthentication techniques such as login using facialrecognition, biometrics or with a passcode.7. Cost-effective for a few mental conditions, such asstress release.8. Provide reminders such as taking medication, doexercise, slots for jogging.

but three therapeutic mental health chatbots[Woebot, Wysa and Tess

The main advantage of these bots is toprovide a practical, evidence-based, and an attractive digitalsolution to help fill the gap of the professional instantly [10].

he mentalhealth professionals have adopted the use of technologyspecifically Artificial Intelligence-based chatbots to addressthe needs of individuals affected by mental health problemsas the first line of defence

social stigma and hesitation [8]

0 per 100,000 mental healthprofessionals are available in high-income nations and 2 per100,000 in low-income nations [5

According to the World Health Organization(WHO), 1 in10 people need mental healthcare worldwide, and differentmental disorders are, portrayed by a combination ofperceptions, feelings, and relationships with others [1].

On the other hand, EAI can help health practitionersto increase the emotional understanding of their patients, thus be-ing able to deliver diagnoses and treatment faster and with moreaccuracy.

speech to text, wherethe audio signal is converted into words, and text sentiment analy-sis, where text is provided with emotional meaning.

One of the most important challenges is tointegrate the technology into the health provision process.

Figure 2: Emotional recognition system overview

Figure 1: Russell’s Circumplex model of emotions

Simplediscrete models associate facial expressions to the basic core of6 emotions recognized universally (namely: anger, fear, surprise,disgust, joy, and sadness), whereas multidimensional approachesparameterize emotions as a lineal combination of different psycho-logical dimensions.

Different codification approaches havebeen proposed in the literature, but one of the most widely used isthe e Facial Action Coding System (FACS) developed by Ekman andFriesen [9]. As a final goal, FACS looks to recognize and categorizeaction units (AUs) which represent the minimum units of muscularactivity that can produce momentary changes in facial appearance.Automatic AU recognition provides as output one or more detectedmicro-movements of facial muscles along with the intensity (froma neutral state) of each AU.

The aim of the workpresented here is to provide an architecture that allows the inte-gration of these technologies into the digital health ecosystemspreviously developed by the authors in [1,2]

Virtual therapists may also encouragepatients to share their thoughts and feelings more deeply than ahuman therapist, as they may be perceived as safer environmentsfor sharing personal information.

personalized therapy.

It can also help doctors to ensure patients are following acertain treatment, as well as assessing the treatment evolution.

EAI can be used to help patients understandtheir emotional state under stressful situations, so they can man-age their emotions and handle difficult situations.

emotional AI-based outreach initiatives at community gathering facilities can help elders more comfortably familiarize with the technology, while indirectly having a positive impact on their attitude toward the technology when later used in the diagnosis and treatment process.

This result implies that policy-wise, an increasing awareness of AI through education and media can produce a positive impact on emotional AI perception in healthcare.

This finding suggests that Japanese elderly patients have negative perceptions of emotional AI-based tools in healthcare in both private and public settings, which is problematic as the target demographic of AI-based healthcare utilities is the older generation in Japan.

Discriminatory concern is also found to have no significant correlation with the attitude toward AI in healthcare.

Kitano (2006) argues the traditional belief of Shintoism predisposes Japanese people to the natural propensity of ascribing the human quality of a heart (心-Kokoro) or spirit to inanimate objects, thus, explaining the embrace of Japan toward robots.

traditionally high degree of trust the Japanese have for authority figures and public healthcare institutions.

In contrast, higher familiarity with AI application in healthcare and medicine was associated with subjects’ positive perception of emotional AI usage in public health facilities

Whereas familiarity with AI in healthcare presents positive correlation with the attitude of private AI use, concern over losing control to AI presents a negative correlation. In other words, those with higher concerns over losing control to AI perceived AI private usage in healthcare more negatively. Surprisingly, concerns for privacy violations and discrimination implicated in the use of AI in a private setting are not significant

robots having emotion is what blurs the line between robots, being perceived as an object or a being, being perceived as artificial or humane, and being perceived with a soul or without a soul.

James Wright's ethnographic study on “Hug”, a mechanical care robot designed to help lift and transfer the elderly on behalf of careers, the author discussed the element of “安心” (anshin—peace of mind) to explain the Japanese caregivers' perception of the mechanical lifting device [21].

The researcher rejected the common argument that empathetic robots should not be considered an alternative to caregivers because their emotions and interactions are artificial and thus not authentic, with two counterpoints. First, based on interviews with the patients and observations of their interactions with the care robot, Aronsson revealed that many patients were factually aware of the differences between care robots and caregivers but chose to interact with them regardless. Second, there were circumstances where the caregivers themselves had to act professionally and provide care for the elderly regardless of their empathy or emotions. In such a context, Aronsson argues that the authenticity of Palro's emotional labor was as valid as the human professional's. Thus, in the context of care facilities, the question of authenticity doesn't translate equally in a philosophical sense, given that empathetic robots can act as effectively in their caregiver's role as their human counterpart.

social robots and humans.

empathetic AI-based chatbots are attracting increasing research attention as they are believed to be beneficial in supporting patients with mental health issues or emotional distress

As mentioned in the previously mentioned, emotional AI is a narrow branch of AI to identify and respond to human emotions. The technology is based on the work of Rosalind Picard, a professor at the MIT Media Lab, who coined the term “affective computing”. Picard defined the concept as “computing that relates to, arises from, and deliberately influences emotion” [12].

Sometimes contradictions result from a conflict between two forces.

For instance, in 1999, Sony introduced the world's first robot dog companion, AIBO, and in 2018, reintroduced AIBO with an AI upgrade that infused it with a “lovable quality”. Studies on the mental health impacts of PALRO, Fujisoft's humanoid robot which communicates with humans through voice and remembers the faces of over 100 people, have demonstrated its ability to reduce anxiety and stress in dementia patients. More importantly, PALRO is found to encourage people with dementia to interact with others in senior care facilities [8].

a shortage of nearly half a million healthcare workers by 2025

emotional AI technologies, i.e., deep learning systems trained to read, classify, and respond to human emotions.