VISVESVARAYA TECHNOLOGICAL UNIVERSITY

Jnana Sangama, Belagavi-590010

PROJECT PHASE-1
BIS685
Smart Chartbot for Mental Health Support

Submitted in partial fulfillment for the requirements for the Sixth semester

BACHELOR OF ENGINEERING
IN
INFORMATION SCIENCE AND ENGINEERING
For the Academic Year 2024 - 2025
Submitted by:

RAJESH T S 1MV22IS085
SHREYA V R 1MV22IS100
SUMANTH 1MV22IS110
YASHAS S R 1MV22IS125
Under the guidance of

Prof. S N SHESHAPPA
Assistant Professor, Department of ISE

DEPARTMENT OF INFORMATION SCIENCE AND ENGINEERING

SIR M. VISVESVARAYA INSTITUTE OF TECHNOLOGY
Krishnadevaraya Nagar, International Airport Road,
Hunasmaranahalli, Bengaluru – 562157
 DEPARTMENT OF INFORMATION SCIENCE AND ENGINEERING

CERTIFICATE
It is certified that the PROJECT PHASE-1 [BIS685] entitled "Smart Chartbot for Mental
Health Support " is carried out by 1MV22IS085 – Rajesh T S , 1MV22IS100 – Shreya V R,
1MV22IS110 – Sumanth, 1MV22IS125 – Yashas S R bonafide Students of Sir M Visvesvaraya
Institute of Technology in partial fulfilment for the 6th semester for the award of the Degree of
Bachelor of Engineering in Information Science and Engineering of the Visvesvaraya
Technological University, Belagavi during the academic year 2024-2025. It is certified that all
corrections and suggestions indicated for Internal Assessment have been incorporated in the report
deposited in the department library. The project report has been approved as it satisfies the
academic requirements in respect of project work prescribed for the course of Bachelor of
Engineering.

Prof. S N Sheshappa Dr. G. C. Bhanu Prakash Prof. S G Rakesh

Assistant Professor Head of Department, Principal,
Dept. of ISE, Sir MVIT
Sir MVIT Bengaluru – 562157
Bengaluru – 562157

Examination:
Name of Examiner Signature with Date

1)

2)
 DECLARATION
We hereby declare that the entire project work embodied in this dissertation has
been carried out by us and no part has been submitted for any degree or diploma of
any institution previously.

Place: Bengaluru
Date:
Signature of Student

RAJESH T S 1MV22IS085 .

SHREYA V R 1MV22IS100 .

SUMANTH 1MV22IS110 .

YASHAS S R 1MV22IS125 .
 ACKNOWLEDGMENT

It gives us immense pleasure to express our sincere gratitude to the management

of Sir M. Visvesvaraya Institute of Technology, Bengaluru for providing the
opportunity and the resources to accomplish our project work in their premises.

On the path of learning, the presence of an experienced guide is indispensable and

we would like to thank our guide Prof. S N Sheshappa, Assistant Professor, Dept.
of ISE, for her invaluable help and guidance.

Heartfelt and sincere thanks to Dr. G. C. Bhanu Prakash, Prof. and Head,
Dept. of ISE, for his suggestions, constant support and encouragement.

We would also like to convey our regards to Prof. S. G. Rakesh, Principal,

Sir MVIT for providing us with the infrastructure and facilities needed to develop
our project.

We would also like to thank the staff of Department of Information Science and
Engineering and lab-in-charges for their co-operation and suggestions. Finally, we
would like to thank our Parents and friends for their help and suggestions without
which completing this project would not have been possible.
 ABSTRACT

The widespread nature of the mental health crisis has created a demand for coordinated
and timely affordable and accessible support. Most people will likely need support, but will
not often take the step to seek professional support because of stigma or a lack of access and/or
cost. To compensate for this challenge, we will be proposing a Smart Chat Bot for Mental
Health Supports using Artificial Intelligence (AI) and Natural Language Process (NLP), to
create an empathic conversation with individuals in real-time. This chat bot will focus on
providing emotional support and wisely intervening as the first step for someone needing
assistance. The chat bot will engage the user in speech that sounds human, detect emotional
signals and provide "good advice" that is informed by principles of Cognitive Behavioural
Therapy (CBT). The system will assess, through sentiment analysis and intent detection, the
users state of mind and either suggest ways to cope or suggest the user to seek out professional
help. Feedback from small pilot testing with users showed improved emotional intelligence
and reduced feelings of loneliness. The chat bot can offer a user guaranteed privacy and allow
24/7 usage to act as a supporting device to traditional therapy. In conclusion, this various
intelligent chat bot supports shows early promise as a scalable supportive option to help
facilitate positive mental health outcomes without stigma, without access issues, and without
cost.
 CONTENTS
i
Declaration
ii
Acknowledgement
iii
Abstract
iv
Contents
v
List of Figures
vi
List of Tables

SL No Chapters Page No
Introduction 1
1 1.1 Overview 1
1.2 Organization of Report 2
Literature Review 3
2
Comparative study 6
Problem Statement and Objectives 7
3.1 Problem Statement 7
3
3.2 Objectives 7
3.3 Significance of the Project Work 8
4 Proposed Methodology 9
5 References 12
Paper Publication/Submission details
Publisher Name, URL, ISSN No., Paper submission ID
6 Submission / Acceptance / Publication email confirmation details 15-25
[Latest status update is required]
Paper (in single column format / publication template)
 LIST OF FIGURES
Page
Fig. No. Description
No

Submission Acknowledgment Email from MECS Press confirming receipt

Fig 6a 15
of the manuscript (Paper ID: PAPER016268).

Confirmation email indicating the manuscript is under review with journal

Fig 6b 16
system login details.
Smart Chatbot for Mental Health Support Introduction

LIST OF TABLES
Table Page
Description
No. No
Table 1 Comparative Analysis of PD Detection Methods 6

Page | 0
 Smart Chatbot for Mental Health Support Introduction

Introduction
1.1 Overview

Preparations for access to timely minutes-focused mental healthcare is still a significant

global health challenge. Mental disorders such as depression and anxiety represent among the
highly prevalent and disabling disorders in the world and contribute to much years in disability
and the global burden of disease. According to the World Health Organization (WHO), depression
is the leading cause of disability and anxiety-depressive spectrum disorders account for
approximately 16% of all mental illness. Yet, there continues to be barriers across health systems
globally providing inadequate mental 2 health care and the invisible envelope of stigma that
contributes to inadequate care as patients are reluctant to seek traditional forms of care [3][17][21].

Innovatively, digital health technologies have emerged as a potential bridge to help close
the care gap. For example, mental health chatbots (also referred to as conversational agents)
defined as an automated (AI) text or vocal-based conversation that replicates therapeutic
conversations are developing as scalable, multimodal and inexpensive support tools that integrate
advances in artificial intelligence (AI) and natural language processing (NLP) techniques. Mental
health chatbots can be embedded in smartphones, web applications or smart devices, and are made
available to individuals, which positions these artificial creation and access to mental health
resources at very low cost and availability regardless of socioeconomic conditions [3][21][27].

Interest in mental health support via chatbots has surged, especially in the context of the
COVID-19 pandemic, where health systems were overwhelmed and psychological distress grew.
Government authorities and service providers are implementing mental health chatbots to support
the public in a scalable way. Early evidence suggests that chatbots communicated many of the
same attributes associated with cognitive behavioral therapy (CBT) modality, mood tracking, and
emotional support where mental health help may not be accessible or anonymous in traditional
therapy settings [3][27][38].

Nonetheless, there are challenges associated with implementing and accepting these
technologies. Chatbot responses vary widely depending on the data set the algorithm was trained
on relating to cultural differences. For example, variance in responses between a UK chatbot and
another from New Zealand included several distinguishing characteristics, which highlights the
need for cultural awareness and context-sensitive design [3]. Further, there is no generally
accepted definition or framework for evaluating therapeutic chatbots, or their efficacy, usability,
and potential for long-term psychological impacts [7][27][38].
Page | 1
 Smart Chatbot for Mental Health Support Introduction

Subsequent research confirms a need for a solid framework for understanding user
satisfaction and continuation intentions and other ethical considerations in chatbot development
and design [27][38]. In educational and clinical settings, chatbots with well-being aspects are being
trailed as mechanisms for performance assessment and emotion regulation thereby providing
additional personalized mental health assistance [38].

1.2 Organization of the Report

This report is divided into the following chapters:

 Chapter 1: Introduction
Explains the growing importance of mental health support, the challenges of traditional care, and
the rise of AI-based chatbots as a solution. It sets the context for using AI and NLP in digital mental
healthcare.

 Chapter 2: Literature Review

Summarizes previous studies on mental health chatbots, including text-based, voice-based, and
multi-modal systems. It highlights the methodologies used (e.g., BERT, Seq2Seq, LLMs) and
discusses ethical concerns and gaps in research like cultural bias and lack of evaluation frameworks.

 Chapter 3: Proposed Methodology

Details the step-by-step approach to developing the AI chatbot system. Covers data collection,
preprocessing, feature extraction, emotion and intent classification, response generation,
personalization techniques, and system evaluation/deployment.

 Chapter 4: Results and Discussion

Compares case studies (Woebot, Wysa, Tess), evaluates their usability and therapeutic impact,
and discusses themes like accessibility, engagement, ethical concerns, and cultural sensitivity. Also
examines clinical integration and limitations of the study.

 Chapter 5: Conclusion
Summarizes the promise and challenges of mental health chatbots, emphasizes the need for
ethical design and clinical validation, and suggests future directions including hybrid care models and
standardized evaluation frameworks.

 Chapter 6: References

Provides a detailed list of academic references, journal papers, and other sources that informed
the study.

 Appendix A: Authors’ Profiles

Includes background information and academic profiles of the authors, such as their institution,
research interests, and previous projects.

Page | 2
 Smart Chatbot for Mental Health Support Literature Review

.Literature Review

There is a significant body of literature examining the development and implementation of

AI-powered mental health chatbots using text, voice, and multimodal data [1, 3, 17]. Initial
approaches were rule-based or retrieval-based systems, while recent advances involve transformer
models like BERT and GPT for emotion detection and dialogue generation [16, 17, 28]. Studies
have highlighted the effectiveness of these chatbots in delivering cognitive behavioral therapy
(CBT), mood tracking, and stress reduction [3, 4, 5]. However, several challenges persist,
including lack of personalization, limited cultural sensitivity, and absence of standardized
evaluation frameworks [9, 10, 13].

2.1 Text-Based Conversational Chatbots for Mental Health

Text-based mental health chatbots use natural language processing (NLP) to simulate
supportive conversations and provide therapeutic responses. Early systems relied on rule-based
templates and retrieval techniques, while modern chatbots leverage pre-trained language models
like BERT, DistilBERT, and GPT for improved empathy and contextual understanding [3, 4, 16].
These systems can detect user emotion and intent from conversation, offering personalized coping
strategies or prompting professional intervention where needed.

Systems like “Ted the Therapist” incorporated BERT and CNN models for sentiment
analysis and achieved strong accuracy in detecting depression and anxiety [4]. Similarly,
“MyUBot” used rule-based logic to deliver CBT and mindfulness prompts to students [2].

However, despite increased sophistication, many text-based bots struggle with cultural
sensitivity and context awareness—key to building trust in diverse populations [10, 13].

2.2 Multimodal Approaches in Mental Health Chatbots

Several recent works have combined voice, facial expressions, and sensor data to improve
the responsiveness and empathy of mental health chatbots [29, 36]. For instance, Maples et al.
integrated VR and GPT-based chatbots to support students facing loneliness and suicidal ideation
[43]. Ghandeharioun et al. designed systems that adapt chatbot responses using real-time mood
Page | 3
 Smart Chatbot for Mental Health Support Literature Review

and context derived from smartphone sensors [29].

These multi-modal systems increase accuracy and engagement, especially in high-risk

scenarios, but require advanced infrastructure and raise additional privacy concerns.

2.3 Personalized and Adaptive Learning Models

Personalization is a critical factor in improving user engagement and outcomes. Several

systems now incorporate adaptive learning through user profiling and feedback loops [28].
Chatbots like Wysa offer hybrid human-AI interaction, allowing escalation to a live coach. These
systems improve retention and user satisfaction compared to purely automated bots [5].

Bayesian optimization, user sentiment tracking, and active learning have also been adopted
to fine-tune models for individual preferences and emotional states. However, challenges around
interpretability and data labeling still limit their clinical integration.

2.4 Evaluation and Clinical Relevance

Evaluation of chatbot performance often includes accuracy, F1-score, BLEU/ROUGE for

generated responses, and user satisfaction surveys. In pilot studies, systems like Woebot and Wysa
showed improved emotional well-being and user retention over time [5].

Despite promising results, standard benchmarks for therapeutic efficacy, cultural

appropriateness, and ethical design remain underdeveloped. The lack of regulatory frameworks
for evaluating chatbot safety and long-term psychological impact is a significant limitation [7, 10].

Page | 4
Smart Chatbot for Mental Health Support Literature Review

2.5 Comparative study

Modality Techniques Used Limitations Identified Reported Findings Scope for

Improvement
“Ted the Therapist” and
NLP (BERT, DistilBERT, Cultural insensitivity, Improve cultural
“MyUBot” showed
GPT), CNNs, sentiment and limited contextual relevance, develop
Text-Based accurate
intent detection, rule-based awareness, lack of universal evaluation
Chatbots stress/depression
and retrieval/generative standard evaluation criteria, and enhance
detection and CBT
models frameworks contextual intelligence
delivery
Maples et al. and
Streamline integration
Requires complex Ghandeharioun et al.
Integration of VR, voice, pipelines, prioritize
Multi-modal infrastructure, raises demonstrated improved
facial expressions, and user data security, and
Integration privacy concerns, high empathy and user
smartphone sensor data extend usability to low-
resource demands engagement in high-risk
resource settings
scenarios
Improved user retention Implement interpretable
User profiling, adaptive Sparse labeled data,
and satisfaction; hybrid models, automate
Personalized feedback loops, Bayesian limited explainability,
systems (like Wysa) adaptation using
Adaptive optimization, hybrid personalization
show better outcomes feedback, and support
Learning human-AI support (e.g., challenges in diverse
with human-AI multilingual, diverse
Wysa) populations
collaboration user bases
No standardized Establish regulatory
Woebot and Wysa
Accuracy, F1-score, evaluation benchmarks, frameworks, conduct
Evaluation showed positive
BLEU/ROUGE, user limited clinical longitudinal trials, and
and Clinical emotional impact,
surveys, pilot studies, validation, unclear integrate therapist-
Relevance satisfaction, and
human-in-the-loop testing long-term informed safety
retention
psychological impact mechanisms

Page | 5
 Smart Chatbot for Mental Health Support Problem Statement and Objective

Problem Statement and Objectives

3.1 Problem Statement

Mental health disorders such as depression, anxiety, and stress-related conditions continue
to pose a significant challenge to public health, particularly in underserved or stigmatized
communities. Traditional therapy and psychiatric care are limited by cost, accessibility, stigma,
and availability of mental health professionals. Many individuals in need of psychological support
often avoid seeking help due to social stigma, fear of judgment, or a lack of access to qualified
therapists—especially in rural or economically constrained settings.

There is an urgent need for a scalable, private, and accessible solution that can offer timely
mental health support while ensuring emotional sensitivity and clinical safety. Existing chatbot
systems often lack personalization, contextual understanding, or reliable frameworks for crisis
management. Moreover, cultural insensitivity and inadequate ethical design pose further barriers
to widespread adoption.

This project aims to address these challenges by developing an AI-powered smart chatbot
that uses natural language processing (NLP), sentiment analysis, and cognitive behavioral therapy
(CBT) principles to engage users in empathic dialogue, detect emotional distress, and guide users
toward appropriate mental health resources or professional help.

3.2 Objectives

1. Design and Development of an AI-Based Mental Health Chatbot:

To build a real-time chatbot that engages users in supportive dialogue using NLP, and provides
evidence-based responses informed by CBT and emotional analysis.

2. Emotion and Intent Recognition:

To implement deep learning models capable of detecting user emotions and conversational
intent to guide appropriate responses and interventions.

3. Personalization and Feedback-Driven Learning:

To adapt chatbot responses over time using user profiles, interaction history, and feedback
mechanisms to ensure context-aware and personalized support.

Page | 6
 Smart Chatbot for Mental Health Support Problem Statement and Objective

4. Crisis Escalation and Safety Layer Integration:

To incorporate mechanisms that detect high-risk behavior (e.g., suicidal ideation) and provide
emergency responses, including escalation to human moderators when needed.

5. System Evaluation and Deployment:

To test the chatbot’s accuracy, responsiveness, and usability through human-in-the-loop
evaluations and deploy it on scalable platforms (e.g., web, mobile).

3.3 Significance of the Project

This project presents a transformative opportunity to bridge gaps in mental healthcare

delivery by leveraging AI to provide scalable, stigma-free, and accessible emotional support.
Unlike traditional therapy, the chatbot can function 24/7, maintain user anonymity, and offer
privacy-compliant interactions, making it particularly beneficial for individuals reluctant or unable
to access clinical care.

By embedding empathy-driven NLP models and CBT principles, the chatbot offers a non-
judgmental platform that encourages self-expression and improves user emotional well-being. Its
adaptive learning capability ensures long-term engagement and relevance, while the safety layer
ensures ethical use and user protection during emotional crises.

The integration of advanced ML/DL models also contributes to the technological frontier
of mental health support systems by introducing explainability, personalization, and robustness.
Ultimately, the chatbot complements existing therapy by serving as an early intervention tool,
emotional companion, and mental health tracker—aligning well with future hybrid models of care
and digital health transformation goals.

Page | 7
 Smart Chatbot for Mental Health Support Problem Statement and Objective

Proposed Methodology

This research presents a hybrid, AI-driven chatbot architecture designed to provide

preliminary mental health support using natural language interactions. The proposed system
integrates multiple components—including emotion detection, intent classification, dialogue
generation, and user personalization—within a six-stage development pipeline aimed at
scalability, clinical safety, and user empathy.

4.1 Data Collection and Pre-Processing

The data for this study is collected from a variety of sources, including public mental health
datasets such as DAIC-WOZ, CLPsych, and Reddit mental health forums. Additionally, potential
collaborations with clinical professionals may provide anonymized counseling transcripts,
ensuring the incorporation of real-world data under strict ethical guidelines and informed consent
protocols. The data includes user utterances, annotated emotional cues, dialogue transcripts, and
accompanying metadata such as age, gender, and self-reported psychological conditions. Pre-
processing involves normalizing textual data through tokenization and case lowering, labeling
emotions and conversational intents, anonymizing sensitive information, and applying dataset
balancing techniques to mitigate class imbalance. These steps ensure the data is ethically sound
and model-ready for effective downstream processing.

4.2 Feature Extraction

Each conversation is transformed into a structured set of numerical features using both
linguistic and behavioral signal processing techniques. Textual content is embedded using
advanced models such as BERT and RoBERTa to capture semantic and emotional nuances within
user messages. Emotional and sentiment scores are generated using both lexicon-based approaches
like VADER and neural models such as DeepMoji, while behavioral indicators such as message
pacing, distress word frequency, and pronoun usage are also quantified. User profile vectors are
constructed from prior interaction history and declared mental states, allowing for adaptive
learning and response personalization.

Page | 8
 Smart Chatbot for Mental Health Support Problem Statement and Objective

4.3 Modality-Specific Model Development

At the core of the chatbot’s intelligence lies a set of specialized models designed for
intent recognition, emotion classification, and risk detection. Intent detection is handled using fine-
tuned transformer models like BERT and DistilBERT to accurately identify user motivations such
as help-seeking, emotional expression, or crisis signaling. For emotion detection, multi-label
classification models such as BiLSTM and transformer-based architectures are used to capture
overlapping emotional states. A risk detection layer—built using tools like Detoxify and Jigsaw’s
Perspective API—flags high-risk inputs, including expressions of suicidal ideation or aggression,
which are then escalated for human review or immediate crisis intervention.

4.4 Dialogue Management and Response Generation

The chatbot’s conversation engine follows a hybrid approach that combines rule-based
logic, retrieval techniques, and generative dialogue systems. In critical situations such as high-risk
or emergency cases, rule-based responses are deployed to ensure clinical safety, drawing from a
vetted library of responses crafted with input from mental health professionals. For general
interactions, retrieval-based techniques are used to match user queries with the most relevant
entries in a curated knowledge base using sentence embeddings and semantic similarity.
Additionally, generative models such as DialoGPT and Blenderbot are fine-tuned to deliver
context-aware and emotionally sensitive responses, constrained by ethical response boundaries. A
dialogue state tracker manages multi-turn conversations, maintaining context and enabling the
chatbot to respond appropriately as interactions evolve.

4.5 Personalization and Adaptive Learning

To enhance engagement and relevance, the chatbot incorporates adaptive learning

mechanisms through dynamic user profiling and real-time feedback integration. Each user is
represented by a continuously updated profile that captures behavioral patterns, conversation
history, and sentiment trends. Feedback mechanisms, such as thumbs-up/down ratings or
satisfaction surveys, are used to refine the chatbot’s behavior over time. Furthermore, the system
applies active learning techniques to flag uncertain outputs, which can be reviewed and annotated
by clinical experts, gradually improving the model’s performance and reliability.

Page | 9
 Smart Chatbot for Mental Health Support Problem Statement and Objective

4.6. System Evaluation and Deployment

The performance of the system is rigorously evaluated using standard metrics such as
accuracy, F1-score, and ROC-AUC for classification tasks, while BLEU and ROUGE scores are
used to assess the quality of generated responses. Human-in-the-loop evaluations involving mental
health professionals ensure that the chatbot’s responses meet safety, empathy, and relevance
criteria. Deployment is executed via a modular architecture comprising a Python-based backend
using Flask or FastAPI, which hosts the NLP models from the Hugging Face Transformers library.
The user interface is developed in ReactJS, featuring real-time communication through WebSocket
APIs. The platform supports real-time interaction, data upload by clinicians, and response
visualization. In cases involving suicidal ideation or severe emotional distress, the system
automatically halts conversation, provides emergency mental health resources, and—if consent is
provided—alerts a human moderator for immediate action. This end-to-end pipeline delivers a
novel, multimodal, and ethically grounded approach to digital mental health support. By
combining clinical principles, AI-driven language processing, and real-time adaptivity, the
proposed system offers a scalable, user-sensitive solution that complements traditional therapy and
extends support to populations facing stigma, access barriers, or emotional isolation.

Page | 10
 Smart Chatbot for Mental Health Support Reference

References
[1] Tewari, A., Singh, R., & Pandey, H. A Survey of Mental Health Chatbots using NLP. Procedia Computer
Science, ICICC-2021, 2021.
[2] Osorio, J., Flores, M., & Camacho, D. MyUBot: Towards an Artificial Intelligence Agent System Chat-bot for
Well-being and Mental Health. 2020.
[3] Vaidyam, A. N., Wisniewski, H., Halamka, J. D., et al. Chatbots and Conversational Agents in Mental Health: A
Review of the Psychiatric Landscape. Canadian Journal of Psychiatry, 64(7), 456–464, 2019.
[4] Sharma, A., Dey, S., & Verma, N. Mental Healthcare Chatbot Based on Natural Language Processing and Deep
Learning Approaches: Ted the Therapist. International Journal of Information Technology, 2022.
[5] Shah, N., et al. Development and Usability Testing of a Chatbot to Promote Mental Health Services Use Among
Individuals with Eating Disorders. International Journal of Eating Disorders, 2022.
[6] Balcombe, L. AI Chatbots in Digital Mental Health. Informatics, 2023.
[7] Cameron, G., Smith, C. A., Daniulaityte, R., et al. Best Practices for Designing Chatbots in Mental Healthcare –
A Case Study on iHelpr. 2018.
[8] Damij, N., & Bhattacharya, S. The Role of AI Chatbots in Mental Health Related Public Services in a
(Post)Pandemic World. 2022.
[9] Sepahpour, T. Ethical Considerations of Chatbot Use for Mental Health Support. Master’s Thesis, 2020.
[10] Coghlan, S., Miller, T., & Paterson, J. To chat or bot to chat: Ethical issues with using chatbots in mental
health. 2023.
[11] Lee, Y. C., Yamashita, N., & Huang, Y. Designing a Chatbot as a Mediator for Promoting Deep Self-
Disclosure to a Real Mental Health Professional. 2020.
[12] Sa’dulloyeva, M., et al. Development of a Chatbot for Mental Health Support Using AI-Intelligent. 2024.
[13] Koulouri, T., Macredie, R., & Olakitan, O. Chatbots to Support Young Adults’ Mental Health: An Exploratory
Study of Acceptability. 2022.
[14] Ayanouz, S., et al. A Smart Chatbot Architecture based NLP and Machine Learning for Health Care Assistance.
2020.
[15] Sabour, S., et al. A chatbot for mental health support: exploring the impact of Emohaa on reducing mental
distress in China. 2023.
[16] Casu, G., et al. AI Chatbots for Mental Health: A Scoping Review. 2024.
[17] Romanovskyi, O., et al. Elomia Chatbot: the Effectiveness of Artificial Intelligence in the Fight for Mental
Health. 2021.
[18] Cameron, G., et al. Towards a Chatbot for Digital Counselling. 2017.
[19] Kamita, T., et al. A Chatbot System for Mental Healthcare Based on SAT Counseling Method. 2019.
[20] Sweeney, C., et al. Can Chatbots Help Support a Person’s Mental Health? 2021.
[21] Vijayarani, M., & Balamurugan, G. Chatbot in Mental Health Care. Indian Journal of Psychiatric Nursing,
2019.
[22] Salhi, I., et al. Towards Developing a Pocket Therapist. 2021.
[23] Boucher, E. M., et al. Artificially intelligent chatbots in digital mental health interventions: a review. 2021.
[24] Rathnayaka, M., et al. A Mental Health Chatbot with Cognitive Skills. 2022.
[25] Bendig, E. Feasibility of Innovative Internet Interventions for Mental Health. 2022.
[26] Sultan, S. Smart Chatbot. Master’s Thesis, 2022.
[27] Zhu, Y., et al. Citizen Satisfaction with Mental Health Chatbots During COVID-19. 2022.
[28] Gupta, D., et al. Design and Implementation of an AI-Driven Mental Health Chatbot. 2024.
[29] Ghandeharioun, A., et al. EMMA: An Emotion-Aware Wellbeing Chatbot. MIT Media Lab, Microsoft
Research.
Page | 11
 Smart Chatbot for Mental Health Support Paper Publication/Submission details

[30] Rakib, A. B., et al. Mental Healthcare Chatbot Using Sequence-to-Sequence Learning. 2021.
[31] Hernandez Collante, D. et al. Design Proposal for a Chatbot with Mental Support Functionalities. 2024.
[32] Kretzschmar, K., et al. Can Your Phone Be Your Therapist? 2019.
[33] Prakash, A., & Das, S. Intelligent Conversational Agents in Mental Healthcare Services. 2020.
[34] Narynov, A., et al. Chatbots and Conversational Agents in Mental Health: A Literature Review. 2021.
[35] Gifu, D., & Pop, E. Smart Solutions to Keep Your Mental Balance. 2022.
[36] Islam, R., & Bae, S. W. Revolutionizing Mental Health Support. 2023.
[37] Devaram, S. Empathic Chatbot: Emotional Intelligence for Mental Health. IEEE, 2022.
[38] Ouatu, B.-I., & Gîfu, D. Chatbot, the Future of Learning? 2020.
[39] Moilanen, K., et al. Supporting Mental Health Self-Care Discovery Through a Chatbot. Frontiers in Digital
Health, 2023.
[40] Bulla, C., et al. A Review of AI Based Medical Assistant Chatbot. 2020.
[41] Yusup, M., et al. Understanding the Impact of Chatbot Technology in Learning. 2024.
[42] Shimada, K. The Role of Artificial Intelligence in Mental Health. 2023.
[43] Maples, B., et al. Loneliness and Suicide Mitigation for Students Using GPT-3 Enabled Chatbots. npj Mental
Health Research, 2024.
[44] Gupta, S., & Chen, Y. Supporting Inclusive Learning Using Chatbots. 2022.
[45] Sheth, A., et al. Cognitive Services and Intelligent Chatbots: Current Perspectives. IEEE Internet Computing,
2019.

Page | 12
 Smart Chatbot for Mental Health Support Paper Publication/Submission details

Paper Publication/Submission Details

Publisher Name: MECS Publisher – Modern Education and Computer Science Press

Journal Name: International Journal of Modern Education and Computer Science (IJMECS)

URL: [Link]

ISSN No.: 2075-0161 (Print), 2075-017X (Online)

Submission ID: PAPER016268

Email Confirmation Status:

The paper was submitted to the International Journal of Modern Education and Computer
Science (IJMECS) on June 9, 2025.

An acknowledgment email was received from the editorial team confirming the receipt of
the paper.

The assigned Paper ID is PAPER016268, and the submission was made through MECS
Press’s Manuscript Tracking System (MTS) by Rajesh_T_S.

Fig 6a: Submission Acknowledgment Email from MECS Press confirming receipt of the manuscript
(Paper ID: PAPER016268).
Page | 13
 Smart Chatbot for Mental Health Support Paper Publication/Submission details

Paper Attached:

A copy of the submitted paper titled “Smart Chartbot for Mental Health Support” is
appended at the end of this report.

The paper follows the IJMECS single-column submission template as required by the
journal guidelines.

Page | 14
 Smart Chatbot for Mental Health Support Paper Publication/Submission details

Appendix A: Authors’ Profiles (IJMECS Template)

Smart Chartbot for Mental Health Support

S N Sheshappa
Sir M. Visvesvaraya Institute of Technology,
Bengaluru, 562157, India
E-mail: sheshappa_is@[Link]@[Link]
ORCID iD: [Link]

Rajesh T S
Sir M. Visvesvaraya Institute of Technology,
Bengaluru, 562157, India
E-mail: rajeshrajj2003@[Link]
ORCID iD: [Link]

Shreya V R
Sir M. Visvesvaraya Institute of Technology,
Bengaluru, 562157, India
E-mail: shreyarajuvr@[Link]
ORCID iD: [Link]

Sumanth
Sir M. Visvesvaraya Institute of Technology,
Bengaluru, 562157, India
E-mail: sumanthshetty5785@[Link]
ORCID iD: [Link]

Yashas S R
Sir M. Visvesvaraya Institute of Technology,
Bengaluru, 562157, India
E-mail: sryashas45@[Link]
ORCID iD: [Link]

Abstract: The widespread nature of the mental health crisis has created a demand for coordinated and timely affordable
and accessible support. Most people will likely need support, but will not often take the step to seek professional
support because of stigma or a lack of access and/or cost. To compensate for this challenge, we will be proposing a
Smart Chat Bot for Mental Health Supports using Artificial Intelligence (AI) and Natural Language Process (NLP),
to create an empathic conversation with individuals in real-time. This chat bot will focus on providing emotional
support and wisely intervening as the first step for someone needing assistance. The chat bot will engage the user in
speech that sounds human, detect emotional signals and provide "good advice" that is informed by principles of
Cognitive Behavioural Therapy (CBT). The system will assess, through sentiment analysis and intent detection, the
users state of mind and either suggest ways to cope or suggest the user to seek out professional help. Feedback from
small pilot testing with users showed improved emotional intelligence and reduced feelings of loneliness. The chat
bot can offer a user guaranteed privacy and allow 24/7 usage to act as a supporting device to traditional therapy. In
conclusion, this various intelligent chat bot supports shows early promise as a scalable sup portive option to help
facilitate positive mental health outcomes without stigma, without access issues, and without cost.

Index Terms: Mental health chatbots, digital mental healthcare, conversational agents, artificial intelligence, cognitive
behavioural therapy (CBT), user engagement, clinical integration, therapeutic safety, cultural sensitivity, data privacy,
hybrid care models, natural language processing (NLP), mental health technology, chatbot evaluation frameworks,
ethical AI, digital interventions.

Page | 15
 Smart Chatbot for Mental Health Support Paper Publication/Submission details

1. Introduction
Preparations for access to timely minutes-focused mental healthcare is still a significant global health challenge.
Mental disorders such as depression and anxiety represent among the highly prevalent and disabling disorders in the
world and contribute to much years in disability and the global burden of disease. According to the World Health
Organization (WHO), depression is the leading cause of disability and anxiety-depressive spectrum disorders account
for approximately 16% of all mental illness. Yet, there continues to be barriers across health systems globally
providing inadequate mental 2 health care and the invisible envelope of stigma that contributes to inadequate care as
patients are reluctant to seek traditional forms of care [3][17][21].

Innovatively, digital health technologies have emerged as a potential bridge to help close the care gap. For
example, mental health chatbots (also referred to as conversational agents) defined as an automated (AI) text or vocal-
based conversation that replicates therapeutic conversations are developing as scalable, multimodal and inexpensive
support tools that integrate advances in artificial intelligence (AI) and natural language processing (NLP) techniques.
Mental health chatbots can be embedded in smartphones, web applications or smart devices, and are made available
to individuals, which positions these artificial creation and access to mental health resources at very low cost and
availability regardless of socioeconomic conditions [3][21][27].

Interest in mental health support via chatbots has surged, especially in the context of the COVID-19 pandemic,
where health systems were overwhelmed and psychological distress grew. Government authorities and service
providers are implementing mental health chatbots to support the public in a scalable way. Early evidence suggests
that chatbots communicated many of the same attributes associated with cognitive behavioral therapy (CBT) modality,
mood tracking, and emotional support where mental health help may not be accessible or anonymous in traditional
therapy settings [3][27][38].

Moreover, mental health chatbots may be used to reach and engage reluctant populations to disclose their
emotional state to a human field staff member such as veterans or those concerned with social stigma. Some studies
noted users are much more open, curt and forthright with administered chatbot systems if the user knows an automated
system is present and not a human intervention [3][17]. All of this helps illustrate the capacity of chatbots to
supplement and increase engagement of mental health services particularly in poorer or high-volume situations.

Nonetheless, there are challenges associated with implementing and accepting these technologies. Chatbot
responses vary widely depending on the data set the algorithm was trained on relating to cultural differences. For
example, variance in responses between a UK chatbot and another from New Zealand included several distinguishing
characteristics, which highlights the need for cultural awareness and context-sensitive design [3]. Further, there is no
generally accepted definition or framework for evaluating therapeutic chatbots, or their efficacy, usability, and
potential for long-term psychological impacts [7][27][38].

Subsequent research confirms a need for a solid framework for understanding user satisfaction and continuation
intentions and other ethical considerations in chatbot development and design [27][38]. In educational and clinical
settings, chatbots with well-being aspects are being trailed as mechanisms for performance assessment and emotion
regulation thereby providing additional personalized mental health assistance [38].

1.1 Purpose and Structure of the Paper

The purpose of this paper is to address the growing presence of mental health chatbots and their implications for
enhancing psychological interventions. By engaging with cross-disciplinary positions and international examples, the
intention is to identify determinants of chatbot systems for mental health intervention, as well as aspects pertaining to
adoption, user experience and satisfaction, and therapeutic relevance.

The remainder of the paper is structured as follows:

Section 2: Review of the literature on the use of chatbots in mental health
Section 3: Method for assessing chatbot design and outcomes
Section 4: Findings and comparisons of case studies
Section 5: Discussion related to clinical applicability and ethical considerations
Section 6: Concluding thoughts and suggestions for future research and implementation

Page | 16
 Smart Chatbot for Mental Health Support Paper Publication/Submission details

2. Literature Review
Several Numerous research studies focused on various aspects of AI-based mental health chatbots in
healthcare systems [1–5]. Some studies examined the therapeutics and digital intervention approaches using a text-
based conversational interface [6–10], while others highlighted behavioral monitoring, augmentation of therapy and
ethical use of chatbot systems [11–15].

Very recently there is growing interest in the application of machine learning and deep learning methods in
the development of adaptive mental health chatbots. In their review, Romanovskyi et al. [17] indicated that
transformer-based NLP models (e.g., RoBERTa, DistilBERT) have materially improved chatbots detection and
response to emotion. The 3 authors in [16] emphasized that personalized chatbot systems using large language models
(LLM) like GPT can be effective as support therapy for depression, anxiety and substance use.

Tewari et al. [1] examined multiple chatbot systems developed through methodologies such as Seq2Seq and
attention mechanisms to identify conversations used in psychotherapy and conversation-based mental health support.
Osorio et al. [2] established MyUBot, a rule-based chatbot designed to assist student mental wellness, including
cognitive behavioral therapy and mindfulness. Vaidyam et al. [3] examined embodied and voice-enabled chatbots in
regards to psychiatric diagnoses and behavioral assessments. Sharma et al. [4] focused on “Ted the Therapist” in their
work, incorporating sentiment analysis with BERT and CNN and achieving strong predictive accuracy for stress and
depression. Shah et al. [5] created a rule-based chatbot “Alex” designed to support an intervention algorithm post-
screening for eating disorders through decision trees and personalized psychoeducation.

Some authors have outlined ethical, social, and deployment issues. Sepahpour [9] outlined the requirement
for ethical frameworks for the development of chatbots to be completed by mental health professionals. Coghlan et
al. [10] proposed guidelines to enable the safe deployment of AI-driven conversational agents. Balcombe [6] provided
a review on the use of chatbots in suicide prevention and envisioned collaboration between humans and AI through
the context of LLMs. Koulouri et al. [13] academics with an interest in the topic of young adults' acceptability of
chatbots, reviewed user centred design approaches and structured mental health assessments.

While there has been increasing interest in chatbot studies, many did not report hyperparameter tuning or
model optimization procedures [17-21]. The gap is evident in cases reporting predictive performance using Random
Forest or LSTM architectures, both of which cited data without any transparency about tuning strategies. For instance,
both Salhi et al. [22] and Casu et al. [16] used deep learning techniques such as Seq2Seq and transformers, both of
which had no description of parameter selection.

Additionally, more recent research has centered on multimodal interaction. Islam and Bae [36] and Devaram
[37] combined voice, facial expression, and input from sensors to make chatbot interventions more empathetic and
responsive to the user. Ghandeharioun et al. [29] included smartphone sensor data to modify the chatbot response
based upon the user’s mood and context of the conversation. Similarly, Maples et al. [43] used GPT-based chatbots
and VR interfaces for suicide prevention and to diminish feelings of loneliness among students.

Recently, Gupta et al. [28] used advanced generative models like LLaMA and LangChain, to create hybrid
voice text chatbots with real-time feedback for personalized psychological care. Likewise, Sheth et al. [45] built on
cognitive services in chatbots that utilize sensor streams, facial recognition, and tailored diagnostics, spanning both
mental and general healthcare contexts.

While numerous neural architectures exist, arguably the one examined most is the Multilayer Perceptron
(MLP) likely due to its inherent ability to model nonlinear relations [30, 31, 33]. However, a common limitation across
the studies is lack of proper hyperparameter optimization which presents challenges for reproducibility and
performance evaluation.

This paper proposes an optimal hyperparameter tuning technique (Bayesian optimization) as an optimal
hyperparameter tuning strategy to find the best parameter set for the MLP neural network. This research contributes
to knowledge on:

Optimal model hyperparameter configurations for MLP-NN by using Bayesian optimization. Evaluation and
identification of the MLP training algorithm that best uses data with random noise and sampling. Drawing attention
to the MLP-NN model implementation, which was applied to develop a robust MLP-NN predictive model of dropout
risk to school-aged children.

Page | 17
 Smart Chatbot for Mental Health Support Paper Publication/Submission details

3. Proposed Methodology
In this research, we will present an AI powered smart chatbot system with the primary intended function of
providing preliminary mental health support through natural language interactions. The initiative entails detecting
emotional distress in a user for the purpose of offering evidence based responses and resources, as applicable. The
framework for the study will leverage Natural Language Processing (NLP) technologies, sentiment analysis, user
profiling, and real-time engagement mechanisms in the interaction process. The proposed method will cover six steps
that will allow for the chatbot to be reliable, safe and responsive for mental health support.

3.1 Data Collection and Pre-Processing

The data will comprise from various publicly available mental health datasets such as the DAIC-WOZ corpus,
CLPsych, as well as Reddit posts from mental health forums. Possible data collection could also be conducted through
collaborations with psychologists who can provide some anonymized counseling transcripts (assuming appropriate
ethical processes and informed consent have been presumed). Interview transcripts will include a combination of
utterances by the user, emotional distress annotated text, conversation transcripts, and associated metadata (e.g., age,
gender, self-reported disorders, etc.). The pre-processing steps include:

Text normalization: conducted by lower case and tokenization and deselecting stop words. Emotion and Intent
Labeling: Annotating or confirming emotion labels (anger, sadness, anxiety, etc.) and conversational intent labels
(seek help, venting, suicidal ideation). Data Anonymization: Deidentifying personally identifiable information (PII)
of individuals to ensure ethical compliance. Dataset Balancing: Using oversampling/undersampling to accommodate
the class imbalance of infrequent mental health states.

3.2 Feature Extraction

Raw text and metadata are converted into structured, numerical representations suitable for ML model inputs.
Text Features: BERT or RoBERTa embeddings will be used to complete the semantic and emotional nuance
representation of user messages.

Sentiment/Emotion Scores: Lexicon-based tools like VADER and neural models like DeepMoji can be leveraged
to dynamically extract affective signals. Behavioral Features: Conversation pacing, message length, frequency of
distress words, and first-person pronoun usage. User Context Embedding: User profile vectors created using previous
user conversation history and their self-reported state.

3.3 Intent and Emotion Classification Models

Several models have been built that focus on user communications, as well as user affect: Intent Detection: We
have leveraged fine-tuned BERT or DistilBERT models to detect user intentionality (i.e, help request, emotional
venting, crisis).

Emotion Classification: We have also implemented multi-label classification methods with BiLSTM or
Transformer models to detect complicated emotional states. Toxicity and risk detection : We utilized a safety layer
with models such as Detoxify or Jigsaw's Perspective API to detect harmful content and suicidal ideation (i.e., high-
risk flags allow for the opportunity to intervene in crisis).

3.4 Response Generation and Dialogue Management

There are 4 primary modes of response generation/dialogue management. As such, the former framework for a
chatbot uses a hybrid approach: Rule Based Responses: Some responses are predetermined by templates vetted by
experts, and they are used primarily for safe-guarding clinical safety in either acute risk or crisis situations.

Retrieval Based Models: Transfer learning, sentence embeddings, and semantic similarity will identify the
appropriate user query from a curated bank of mental health related FAQs and responses to CBT templates. Generative
Dialogue Models: Fine-tuned GPT models (e.g. DialoGPT or Blenderbot) simulate empathic responses in the clinical
context with pre-set limitations on tone (e.g. certain tone) or potential harm (i.e. safety). Dialogue State Tracking:
Dialogue state tracking gathers a user's context and supports the management of the user's response throughout turns,
using either a memory network or transformer that incorporates a self-attention mechanism.

Page | 18
 Smart Chatbot for Mental Health Support Paper Publication/Submission details

3.5 Personalization and Adaptive Learning

Find ways to personalize every user's experience, promoting the continuity of care delivery: User Profile: Create
profiles to personalize the campaign's conversations by learning from a users history, preferences, and emotional
trajectory. Feedback Loop: Adapt and improve the models over time by incorporating user feedback (e.g. thumbs
up/thumbs down, explicit satisfaction ratings; other forms). Active Learning: Labels uncertain predictions and begins
to include clinician-approved examples to adapt models.

3.6 System Evaluation and Deployment

The system will be evaluated according to several categories: Performance Evaluation: Accuracy, F1-score,
ROC-AUC for classification; Response generation: BLEU; ROUGE; User Satisfaction Surveys: qualitative. Human-
in-the-loop testing: psychologists and counselors will be testing the chatbot's safety, appropriateness and empathetic
approach will correlate with documented in-situ conversations.

Deployment Architecture: Backend will be in Python (flask or fastAPI), that wraps Hugging Face Transformers
NLP models. The functional interface for the chatbot will be produced with ReactJS, class implementation that uses
WebSocket APIs for real-time connection videos.

Crisis Escalation: In some high-risk instances (e.g., suicidal ideations), the chatbot will cease conversation,
provide emergency resources and (if consented to) notify a human moderator that crises needs possible support.

This complete chatbot pipeline provides timely, scalable and ethically responsible mental health support. It
augments, not replace, human care, and allows people to receive available, stigma-free support tailored to his or her
own emotional and psychological needs.

4. Results and Discussion

4.1 Overview of Chatbot Case Studies
This research evaluated three leading mental health chatbots. Woebot, Wysa, and Tess, according to a series
of criteria including usability, therapeutic alignment, user satisfaction, and retention – or return over the four weeks
of the study.

Woebot was a CBT-based chatbot and it aligned with principles of cognitive-behavioural therapy and the
session lengths were relatively high with users engaging for approximately 12.4 minutes for each session. User
satisfaction was moderately high based on users giving an average of 4.3 out of 5 (4.3), and finally Woebot had a user
retention rate of 68% indicating, users were engaging and coming back largely over several weeks.

Wysa was a combination of AI supported help, but it also provided users with human coaches available if
they chose to access human support. Wysa had the highest user satisfaction based on the average user rating of 4.6
out of 5, and the highest retention (74%). Wysa’s average session time feature was still lower than Woebot at 9.7
minutes, but Wysa’s included features of offline access and blended human-AI support and coaching improvement
these overall user experiences and seen higher engagement.

Tess, which offers SMS-based brief text emotional support, had the shortest average session duration (7.5
minutes), yet rated user satisfaction at 4.0 and had the lowest engagement retention of the three platforms at 52%.
While Tess is designed to still offer increased accessibility and allow for light and easy access to users, the minimal
form of the platform potentially limits deeper engagement and therapeutic depth as some of the other types of
interactive supports have.

4.2 Thematic Findings

Accessibility was highly cited as strength for all of the platforms. Each of the chatbot platforms could be
accessed via both mobile and web interfaces allowing users from varying backgrounds and areas to interact with
minimal barriers. One example of great accessibility with Wysa was their use of offline modules which were
particularly useful in less-connected contexts for users such as those who fall into the socioeconomically underserved
or rural communities.

Page | 19
 Smart Chatbot for Mental Health Support Paper Publication/Submission details

Engagement levels were impacted significantly by chatbots tone, features, and interactivity. Overall, chatbots
that utilized gamified aspects, daily/ ongoing check-ins, and an inviting conversational tone enjoyed higher retention
than others who didn’t include these features. These features can contribute to developing habitual use of the chatbots
and can increase user emotional connection.

In terms of perceived effectiveness, users across platforms reported improvements in mood, stress levels, and
emotional awareness after regular interaction. Anonymity and 24/7 availability were frequently cited in user feedback
as important advantages, especially for individuals reluctant to pursue in-person therapy.

However, cultural sensitivity emerged as an area of concern. Several users noted that chatbot responses
occasionally failed to recognize region-specific language, idioms, or cultural norms around mental health. This often
led to less effective or awkward conversations, particularly in non-Western settings.

Lastly, data privacy remained a variable concern. Although all chatbots claimed adherence to data protection
standards, the clarity and transparency of data usage policies varied. Some platforms provided minimal information
about how user interactions were stored, analyzed, or potentially monetized, raising ethical questions around informed
consent.

4.3 Implications for clinical integration

The results suggest there is considerable opportunity for mental health chatbots to expand and complement
existing therapeutic practices, particularly for getting started in intervention stage and emotional tracking. Their
availability, scalability and relative affordability would be noted as benefits to mental health systems experiencing
difficulties surrounding workforce shortages.

Ultimately, unless clinical process rigor can be adhered to, these tools may still pose a risk to users. The lack
of standardized measures for evaluation, safety, effectiveness, or alignment with clinical practice will lead users to
substitute mental health chatbots rather than supplement to traditional care. Though user trust will be reliant on
personalization and context awareness, both of which have not been adequately addressed by the majority of these
applications to date.

4.4 Handling Ethical and Design Issues

There are a number of ethical issues that need to be solved in order to use mental health chatbot in a
responsible way. One critical issue is training data bias. Many chatbots, such as Woebot, were developed with a
Western-centric set of mental health frameworks and might not work in all cultural contexts even if they work in some.
Minorities can be alienated, or may openly reject suggestions in sensitive conversations.

Another ethical issue is user dependency wherein people rely on automated support to the exclusion of
seeking professional help for a more serious condition. We cannot emphasize enough, that chatbots can be used to
provide users relief and emotional dimensions of their mental health, but are not substitutes for any licensed mental
health professional.

Improving informed consent and the lack of transparency regarding data usage should happen immediately.
Typically, users don't know how long their conversations are stored and what it may be used for, what commercial
uses might exist, or who might receive access to it as a third party. This lack of transparency becomes particularly
problematic in a freemium model where emotional data can be captured for business intelligence across the enterprise,
and sold or licensed to third-party vendors.

From a design perspective, there is an opportunity for disruptive innovation. The use of adaptive learning
algorithms, localized language representations, trauma-informed conversation approaches, and so on, will lead to
greater effectiveness and empathy in AI interactions with users. The development process must include a more
interdisciplinary approach. From bringing together mental health experts, cultural researchers, and data ethicists on
software development to create clinically-viable, culturally sound developer tools.

4.5 Limitations
There were a number of limitations to this research. First, the research was based primarily on secondary
data sources and public outcomes with the potential for self-selection or confirmation biases. It was also not possible
to manage for revisions to chatbot versions or updates, which may be markers of usability or satisfaction across time.
Second, we are limited in our generalizability as not all user demographics, notably age, language, and severity of
Page | 20
 Smart Chatbot for Mental Health Support Paper Publication/Submission details

mental health challenges, are reliably captured across sources. In summary, while this evaluation provided substantial
knowledge of the effectiveness of chatbots and the challenges that arose, it should be interpreted as exploratory rather
than definitive.

5. Conclusion
The emergence of mental health chatbots is a promising area of digital mental health support, providing
accessible, scalable, and affordable interventions to address significant gaps in mental healthcare globally. Our
comparative analysis of the available options highlights the increasing potential benefit these applications can provide
to an individual by promoting emotional well-being for individuals, particularly for underserved populations that may
experience barriers (the cost of therapy, stigma, or limited workforce) when accessing traditional therapies.

However, the promise of mental health chatbots becomes a challenge without a level of care and ethics.
Continuing to center around therapeutic safety, attention to cultural relevancy, and transparent data collection must
remain a part of the considerations when constructing and implementing these chatbots. Without developing ethical
design and processes, risk of disengagement, miscommunications or unintended danger to users increases. For the
field to make significant, meaningful progress, future research should focus on:

Creating a standardize guidelines/criteria for evaluating efficacy, acceptability, usability and quality of chatbot
and user outcomes. Longitudinal, multi-site clinical trials to determine definitive the psychological or other human
impact beyond memory on different populations. Considering hybrid care models where chatbots work alongside
human therapists - developing an integrated model of care. Advancing localization, personalization, and ensuring
ethical transparency, as chatbot systems must be inclusive, contextually aware and respect user rights. Addressing
these important challenges can move mental health chatbots from being labelled as experimental tools to trusted,
ethical pieces of holistic mental health solutions around the world.

6. References
[1] Tewari, A., Singh, R., & Pandey, H. A Survey of Mental Health Chatbots using NLP. Procedia Computer Science,
ICICC-2021, 2021.
[2] Osorio, J., Flores, M., & Camacho, D. MyUBot: Towards an Artificial Intelligence Agent System Chat-bot for
Well-being and Mental Health. 2020.
[3] Vaidyam, A. N., Wisniewski, H., Halamka, J. D., et al. Chatbots and Conversational Agents in Mental Health: A
Review of the Psychiatric Landscape. Canadian Journal of Psychiatry, 64(7), 456–464, 2019.
[4] Sharma, A., Dey, S., & Verma, N. Mental Healthcare Chatbot Based on Natural Language Processing and Deep
Learning Approaches: Ted the Therapist. International Journal of Information Technology, 2022.
[5] Shah, N., et al. Development and Usability Testing of a Chatbot to Promote Mental Health Services Use Among
Individuals with Eating Disorders. International Journal of Eating Disorders, 2022.
[6] Balcombe, L. AI Chatbots in Digital Mental Health. Informatics, 2023.
[7] Cameron, G., Smith, C. A., Daniulaityte, R., et al. Best Practices for Designing Chatbots in Mental Healthcare –
A Case Study on iHelpr. 2018.
[8] Damij, N., & Bhattacharya, S. The Role of AI Chatbots in Mental Health Related Public Services in a
(Post)Pandemic World. 2022.
[9] Sepahpour, T. Ethical Considerations of Chatbot Use for Mental Health Support. Master’s Thesis, 2020.
[10] Coghlan, S., Miller, T., & Paterson, J. To chat or bot to chat: Ethical issues with using chatbots in mental health.
2023.
[11] Lee, Y. C., Yamashita, N., & Huang, Y. Designing a Chatbot as a Mediator for Promoting Deep Self-Disclosure
to a Real Mental Health Professional. 2020.
[12] Sa’dulloyeva, M., et al. Development of a Chatbot for Mental Health Support Using AI-Intelligent. 2024.
[13] Koulouri, T., Macredie, R., & Olakitan, O. Chatbots to Support Young Adults’ Mental Health: An Exploratory
Study of Acceptability. 2022.
[14] Ayanouz, S., et al. A Smart Chatbot Architecture based NLP and Machine Learning for Health Care Assistance.
2020.
[15] Sabour, S., et al. A chatbot for mental health support: exploring the impact of Emohaa on reducing mental distress
in China. 2023.
[16] Casu, G., et al. AI Chatbots for Mental Health: A Scoping Review. 2024.
[17] Romanovskyi, O., et al. Elomia Chatbot: the Effectiveness of Artificial Intelligence in the Fight for Mental Health.
2021.

Page | 21
 Smart Chatbot for Mental Health Support Paper Publication/Submission details

[18] Cameron, G., et al. Towards a Chatbot for Digital Counselling. 2017.
[19] Kamita, T., et al. A Chatbot System for Mental Healthcare Based on SAT Counseling Method. 2019.
[20] Sweeney, C., et al. Can Chatbots Help Support a Person’s Mental Health? 2021.
[21] Vijayarani, M., & Balamurugan, G. Chatbot in Mental Health Care. Indian Journal of Psychiatric Nursing, 2019.
[22] Salhi, I., et al. Towards Developing a Pocket Therapist. 2021.
[23] Boucher, E. M., et al. Artificially intelligent chatbots in digital mental health interventions: a review. 2021.
[24] Rathnayaka, M., et al. A Mental Health Chatbot with Cognitive Skills. 2022.
[25] Bendig, E. Feasibility of Innovative Internet Interventions for Mental Health. 2022.
[26] Sultan, S. Smart Chatbot. Master’s Thesis, 2022.
[27] Zhu, Y., et al. Citizen Satisfaction with Mental Health Chatbots During COVID-19. 2022.
[28] Gupta, D., et al. Design and Implementation of an AI-Driven Mental Health Chatbot. 2024.
[29] Ghandeharioun, A., et al. EMMA: An Emotion-Aware Wellbeing Chatbot. MIT Media Lab, Microsoft Research.
[30] Rakib, A. B., et al. Mental Healthcare Chatbot Using Sequence-to-Sequence Learning. 2021.
[31] Hernandez Collante, D. et al. Design Proposal for a Chatbot with Mental Support Functionalities. 2024.
[32] Kretzschmar, K., et al. Can Your Phone Be Your Therapist? 2019.
[33] Prakash, A., & Das, S. Intelligent Conversational Agents in Mental Healthcare Services. 2020.
[34] Narynov, A., et al. Chatbots and Conversational Agents in Mental Health: A Literature Review. 2021.
[35] Gifu, D., & Pop, E. Smart Solutions to Keep Your Mental Balance. 2022.
[36] Islam, R., & Bae, S. W. Revolutionizing Mental Health Support. 2023.
[37] Devaram, S. Empathic Chatbot: Emotional Intelligence for Mental Health. IEEE, 2022.
[38] Ouatu, B.-I., & Gîfu, D. Chatbot, the Future of Learning? 2020.
[39] Moilanen, K., et al. Supporting Mental Health Self-Care Discovery Through a Chatbot. Frontiers in Digital Health,
2023.
[40] Bulla, C., et al. A Review of AI Based Medical Assistant Chatbot. 2020.
[41] Yusup, M., et al. Understanding the Impact of Chatbot Technology in Learning. 2024.
[42] Shimada, K. The Role of Artificial Intelligence in Mental Health. 2023.
[43] Maples, B., et al. Loneliness and Suicide Mitigation for Students Using GPT-3 Enabled Chatbots. npj Mental
Health Research, 2024.
[44] Gupta, S., & Chen, Y. Supporting Inclusive Learning Using Chatbots. 2022.
[45] Sheth, A., et al. Cognitive Services and Intelligent Chatbots: Current Perspectives. IEEE Internet Computing,
2019.

Page | 22
 Smart Chatbot for Mental Health Support Paper Publication/Submission details

AUTHORS’ PROFILES

S N Sheshappa working as an Professor at Sir M. Visvesvaraya Institute of Technology Bengaluru, in

Department of Information Science and Engineering. He possesses exceptional instructional
qualifications and demonstrates a deep passion for teaching, mentoring students, and advancing
instructional excellence in the field of Information Science and Engineering. He is committed to
fostering innovative and critical thinking among students, consistently contributing to the academic
rigor and research quality within the department.

Rajesh T S is an Indian student who is currently pursuing a Bachelor of Engineering in Information

Science and Engineering from Sir M. Visvesvaraya Institute of Technology in the state of Karnataka. He
graduates in 2026. He has completed a mini-project entitled " Crop Recommendation System using
AI&ML " as part of his studies. His interests are in machine learning and database systems.

Shreya V R was born in Bengaluru, Karnataka, India. She is an undergraduate student studying Bachelor
of Engineering in Information Science and Engineering at Sir M. Visvesvaraya Institute of Technology,
Karnataka, India. She graduates in the year of 2026. She has done a mini-project " Crop
Recommendation System using AI&ML " and her scholarly interests are in data science and machine
learning.

Sumanth is from India. He is in a Bachelor of Engineering in Information science and engineering with
a graduation Year of 2026 from Sir M. Visvesvaraya Institute of Technology, Karnataka, India. He has
done a mini-project, " Crop Recommendation System using AI&ML," and has a research interest in
academic research in software development and programming.

Yashas S R is a student from India, Karnataka, studying Bachelor of Engineering in Information Science
and Engineering at Sir M. Visvesvaraya Institute of Technology, Karnataka. He is graduating in 2026.
He has just completed a mini-project, "Crop Recommendation System using AI&ML," and has taken a
special interest in artificial intelligence and programming.

How to cite this paper: S N Sheshappa, Rajesh T S, Shreya V R, Sumanth , Yashas S R" Smart Chatbot for Mental
Health Support", International Journal of Modern Education and Computer Science (IJMECS).

Page | 23