Bridging the Communication Gap: Utilizing Large Language Models to Detect Emotional Distress and Depression in Adolescent Communication for Parental Support

In the digital age, the communication gap between parents and adolescents has increased, presenting challenges to understanding the emotional well-being of their children. With the increasing prevalence of social networks, adolescents tend to express their feelings and struggles online rather than engage in face-to-face interaction. Existing monitoring tools allow parents to read messages and observe social media activity, but often fail to interpret the emotional content. Recent studies have explored the feasibility of using natural language processing and machine learning to predict depression based on social media activity. By analyzing the linguistic patterns, sentiments, and emotional content of online communication, researchers have demonstrated the potential to identify individuals suffering from depression at an early stage.

This article proposes a novel solution that uses large language models (LLMs) to monitor and analyze adolescent communication on digital platforms, including smartphones and social media. The system aims to detect emotional distress, signs of depression, and other mental health indicators, providing timely alerts to parents. This technology enables parents to understand their teens’ emotions, offer the necessary support, and prevent the escalation of anxiety and depression.

Introduction

The advent of the digital age has transformed the way adolescents communicate, primarily through smartphones and social media platforms. Although these technologies have provided new avenues for self-expression and social interaction, they have also created a significant communication gap between parents and their children. This gap makes it increasingly difficult for parents to discern their children’s emotional states and respond appropriately to their needs [1,2].

Adolescence is a critical period for emotional and psychological development. Early detection of emotional distress and depression is essential to prevent the escalation of these issues to more serious mental health problems [3-10]. Traditional monitoring tools, which allow parents to read messages and observe social media activity, often fail to interpret the emotional content behind the communication. As a result, parents can miss vital signs of their children’s emotional struggles. In addition, the increasing use of emojis in adolescent communication adds another layer of complexity [11]. Emojis, which can convey a wide range of emotions and sentiments, are frequently used by teens to express feelings in a subtle and nuanced way [12-13]. This visual language can be challenging for parents to interpret accurately without the aid of advanced tools.

Large Language Models (LLMs) offer a promising solution to these challenges. Using advanced natural language processing capabilities, LLMs can analyze large amounts of text data [14], including emojis, to detect subtle emotional signals and indicators of mental health issues. This paper proposes a novel system that utilizes LLMs to monitor and analyze adolescent communication on digital platforms. The system aims to detect signs of emotional distress, depression, and other mental health indicators, providing timely alerts to parents. This technology enables parents to better understand the emotions of their teens, provide the necessary support, and prevent the escalation of anxiety and depression.

In the following sections, we will explore the methodology of the proposed system, detailing how it monitors communication and detects emotional states. We will also discuss the implementation process, including the design of a user-friendly interface for parents. To illustrate the effectiveness of the system, we will present case studies. Finally, we will address potential challenges and limitations, emphasizing the importance of balancing monitoring with respect to adolescent privacy, and suggest future directions for this innovative approach to supporting adolescent mental health.

Methodology

The proposed system uses Large Language Models (LLMs) to monitor and analyze adolescent communication on smartphones and social media platforms. By integrating with these digital platforms, the system can access text messages, social media posts, and other forms of digital communication, including the use of emojis. The system operates in real-time, continuously scanning for signs of emotional distress and depression [15-17].

Emotion detection techniques

LLMs are trained to detect a wide range of emotions and mental health indicators from text data. The process involves several key techniques: [18-20] Sensitivity analysis: The LLMs analyze the general sentiment of the communication, categorizing it as positive, negative, or neutral.

Natural Language Processing (NLP) Advanced NLP algorithms parse and understand the context of communication. These algorithms identify key phrases and sentiments that indicate emotional states.

Sentiment Analysis Sentiment analysis tools evaluate the tone and mood of the messages [21]. This involves classifying the text as positive, negative, or neutral and detecting subtle variations that can indicate distress [22].

Emoji Interpretation Emojis play a crucial role in modern digital communication [23,24]. The system includes a comprehensive emoji interpretation module that deciphers the emotional context of the emojis used in messages [25].

Contextual Analysis The system considers the context in which messages are sent. This includes analyzing the conversation history, frequency of communication, and changes in communication patterns that might signal emotional issues [25].

Mental health indicators

The system is designed to detect specific indicators of mental health concerns, including

• Depression: Indicators such as expressions of hopelessness, persistent sadness, or loss of interest in activities.
• Anxiety: Signs of excessive worry, fear, or stress.
• Emotional distress: General indicators of emotional turmoil, such as irritability, frustration, or anger.

Privacy and ethical considerations

Ensuring the privacy and ethical handling of adolescent communication data is paramount. The system incorporates several measures to address these concerns:

• Data anonymization: Personal identifiers are removed from data to protect the privacy of adolescents.
• Consent management: The system requires explicit consent from both parents and adolescents before monitoring begins. The consent forms outline the scope of monitoring and data usage.
• Datasecurity: Robust encryption protocols are implemented to secure data during transmission and storage.
• Ethical guidelines: The system adheres to ethical guidelines for monitoring, ensuring that the main focus remains on the well-being and privacy of adolescents. Regular audits are conducted to ensure compliance with ethical standards.
• Parental alerts: Alerts to parents are designed to be informative but sensitive, providing enough information to prompt appropriate support without infringing on adolescent privacy.

System overview

The implementation of the LLM system involves careful integration with digital platforms, the design of an intuitive user interface for parents, and practical examples that demonstrate the effectiveness of the system. This comprehensive approach ensures that parents are well-equipped to understand and respond to their adolescent’s emotional needs, fostering a supportive and communicative environment. The system consists of two distinct parts: a monitoring program and a user interface.

Monitor

The monitoring program is responsible for processing information from the user’s communication on smartphones and social media platforms. Using advanced Natural Language Processing (NLP) algorithms and Large Language Models (LLM), the program scans text messages, social media posts, and emojis in real-time to detect emotional cues and mental health indicators. This process includes sentiment analysis to determine the tone and mood of the messages, contextual analysis to understand the broader context of the communication, and emoji interpretation to decipher the emotional content of visual symbols. By continuously analyzing these data streams, the monitoring program can identify patterns and changes in emotional states, providing a comprehensive understanding of adolescent mental health.

User interface

The second part of the system is the user interface, designed to present processed information to parents in an accessible and actionable way. This interface features a dashboard that visually represents the emotional trends of adolescents over time through graphs and charts, highlighting significant patterns and changes. Real-time alerts and notifications are prominently displayed, providing detailed information about detected indicators of emotional distress or depression. In addition, the interface includes detailed reports that break down the detected emotions, explain the context and frequency, and offer information on the emotional importance of the emojis used in communication. Actionable insights and recommendations are provided to guide parents on how to effectively support their adolescent, including conversation starters, mental health resources, and tips for fostering open communication. Privacy controls within the interface allow parents to manage consent options and ensure transparency and control over the monitoring process, maintaining a balance between effective support and respect for the adolescent’s privacy.

Integration with digital platform

To demonstrate the effectiveness of the proposed system, we present several use cases that illustrate its implementation and immaturity in the real world. These examples show how the system is deployed in practical scenarios, highlighting its capabilities in detecting emotional patterns, providing timely alerts, and enabling parental intervention to support adolescents; mental well-being.

The real-time monitoring system grants access to the messages exchanged according to a specific API and will autonomously provide the results of the analysis to the teen guardian.

The integration of the LLM system into existing digital platforms involves several key steps to ensure seamless operation and effective monitoring.

• API development
• Platform-specific APIs: The system integrates with social networks and messaging platforms through the development of custom APIs. These APIs are customized for each specific platform, allowing the system to access and analyze data from users’ communications. Using these APIs, the system can monitor language and emoji usage across various social media channels, detect emotional patterns, and provide timely alerts to parents about potential mental health concerns.
• Standardized data format: Data from different platforms are standardized to ensure consistent processing and analysis by the LLM system.
• Real-time data processing
• Streaming data integration: The system is designed to process data in real-time, allowing immediate detection and response to indicators of emotional distress.
• Scalability The architecture supports scalability to handle large volumes of data from multiple users simultaneously.
• Secure data handling
• Encryption: When necessary, the resulting data is encrypted to ensure privacy and security.
• Compliance: The system adheres to data protection regulations, such as GDPR and COPPA, to protect user privacy.

Presentation of user interface

• Dashboard
• Overview of emotional trends: A visual representation of the emotional state of an adolescent over time, using graphs and charts to highlight patterns and changes.
• Alerts and notifications: Real-time alerts are prominently displayed, with detailed information about detected emotional distress or depression indicators.
• Detailed reports
• Emotion analysis: In-depth reports provide a breakdown of emotions detected in adolescent communication, including context and frequency.
• Emoji interpretation: A section dedicated to explaining the emotional context of emojis used in messages, helping parents understand their significance.
• Actionable insights
• Recommendations: The system offers suggestions on how parents can support their adolescents based on the detected emotional state. This includes conversation starters, resources for mental health support, and tips for fostering open communication.
• Privacy controls: Parents can manage privacy settings and consent options directly through the interface, ensuring transparency and control over the monitoring process.

The system is designed to provide meaningful emotional insight to parents while preserving the privacy of adolescents. By focusing on emotional trends and actionable insights, the system helps parents support their children effectively without compromising their autonomy and privacy.

System benefits

The following examples illustrate how the LLM system helps parents understand and respond to their teen’s emotional state. By detecting patterns in language and emoji usage, the system provides timely alerts, allowing parents to intervene and support their children effectively. These use cases highlight the system’s ability to recognize signs of anxiety, monitor emotional changes, and ensure overall emotional well-being in adolescents.

• Early detection of depression
• Scenario: The system detects a consistent pattern of negative sentiment and expressions of hopelessness in adolescent messages.
• Scenario: The system detects a consistent pattern of negative sentiment and expressions of hopelessness in adolescent messages.
• Response: An alert is sent to the parent, highlighting potential signs of depression and providing resources to seek professional help. The parent can initiate a supportive conversation, guided by the system recommendations.
• Addressing anxiety
• Scenario: Frequent mentions of worry and stress are identified in the adolescent’s communication, along with a marked increase in anxious emojis.
• Response: The system alerts the parent to signs of anxiety and suggests ways to alleviate stress, such as engaging in calming activities or discussing coping strategies. The parent can use the information provided to better understand and address the concerns of their child.
• Monitoring emotional well-being
• Scenario: The system tracks the emotional trends over time, noticing a sudden shift from positive to negative emotions.
• Response: The parent receives a report that highlights this change, prompting him to check in with their adolescent and offer support. The system recommendations help the parent navigate the conversation and provide the necessary reassurance.

Use cases

In the digital age, monitoring adolescents’ online communications can reveal crucial insights into their mental health. This document presents three scenarios where a system using Large Language Models (LLMs) identifies and addresses emotional issues. By detecting patterns in language and emoji use, the system alerts parents to potential concerns, allowing for timely interventions. These scenarios illustrate the system’s effectiveness in recognizing anxiety in a young athlete, monitoring emotional changes in a teenager, and ensuring overall emotional well-being.

• Scenario 1: Early Detection of Depression
• Background: Emma, a 15-year-old high school student, began exhibiting signs of withdrawal and sadness. Her parents were unaware of the extent of her emotional distress.
• System intervention: The LLM system detected frequent expressions of hopelessness and sadness in Emma’s text messages and social media posts over several weeks. Emojis conveying negative emotions, such as tears and frowns, were also noted.
• Outcome: An alert was sent to Emma’s parents, detailing the emotional indicators and suggesting professional mental health resources. The parents initiated a supportive conversation with Emma, guided by the system’s recommendations. Emma began seeing a counselor, and her emotional state improved over time.
• Scenario 2: Addressing anxiety
• Background: Jake, a 16-year-old athlete, started showing signs of anxiety, especially before competitions. His communication was filled with expressions of worry and stress.
• System intervention: The system identified a pattern of anxious language and emojis indicating nervousness. A significant increase in messages about stress and fear of failure was noted.
• Outcome: Jake’s parents received an alert highlighting the signs of anxiety. They used the system’s recommendations to talk to Jake about his feelings and explore relaxation techniques. Jake began practicing mindfulness exercises, which helped reduce his anxiety.
• Scenario 3: Monitoring emotional well-being
• Background: Lily, a 14-year-old, experienced a sudden shift in her emotional state, going from generally positive to increasingly negative over a short period.
• System intervention: The LLM system tracked the change in Lily’s communication patterns, noting an increase in negative sentiment and changes in emoji usage. The system alerted her parents to this sudden change.
• Outcome: The parents received a detailed report and started a conversation with Lily. They discovered that she was being bullied at school. With the guidance of the system, they took steps to address bullying and support Lily emotionally.

Recommendations and limitations

The implementation of large language models to detect emotional distress and depression in adolescent communication presents a promising approach to bridging the communication gap between parents and their children. However, for this technology to be truly effective and ethically sound, it is essential to consider various recommendations and limitations.

Key recommendations for stakeholders involved in the development and use of this technology include emphasizing the importance of ethical considerations, collaboration, and continuous improvement. Ethical considerations must be at the forefront, ensuring the protection of adolescents’ privacy, minimizing the potential for biased or inaccurate predictions, and mitigating any negative emotional impacts. Collaboration between researchers, mental health professionals, and technology developers is crucial to leverage diverse expertise and perspectives, ultimately leading to more robust and responsible solutions [26,27].

The participation of mental health professionals is crucial to guide the use of this tool. Experts can provide valuable information on the ethical use of technology, ensuring that it aligns with best practices in mental health care. They can assist in developing protocols for identifying and responding to signs of emotional distress and depression. They can help design training programs for parents and educators on how to use the tool effectively and responsibly.

In addition, continuous improvement of the models is necessary to address limitations related to the impact of informal language, such as emoticons and slang, on the performance of sentiment analysis [26]. Ongoing validation and refinement of the models, in partnership with end users, will be essential to improve the accuracy and reliability of emotional distress detection, thus increasing the potential for early intervention and support for adolescents in need.

Although the application of large-language models to this domain holds promise, potential limitations must also be acknowledged. Concerns about the accuracy and generalization of the models, particularly in the context of diverse adolescent populations, must be carefully examined [27]. Furthermore, the potential for bias, both in the training data and the model itself should be thoroughly investigated and mitigated to ensure equitable and inclusive outcomes [27].

The emotional impact on adolescents is another crucial consideration, as the deployment of this technology can inadvertently exacerbate existing stigma or introduce new challenges related to privacy and autonomy [27].

By acknowledging and addressing these recommendations and limitations, the proposed technology can become a more robust, ethical, and effective tool to support adolescent mental health.

Conclusion

In conclusion, leveraging large language models to monitor adolescent communication offers a promising approach to bridging the communication gap between parents and their children. By detecting early signs of emotional distress and depression, the system allows parents to provide timely and effective support. This technology not only enhances parental understanding of their teens’ emotional states but also plays a crucial role in promoting mental health and well-being in the digital age.

The benefits of using LLMs extend beyond mere detection. They foster better communication, provide actionable information, and ultimately contribute to a supportive environment where adolescents feel understood and cared for. As technology continues to advance, the potential for further improvements and innovations in this field is great, paving the way for a more effective and empathetic mental health support system. For optimal effectiveness, the system should involve professional psychologists in its evaluation process. Their expertise can help fine-tune the model, ensuring more accurate detection of emotional distress and more precise recommendations for parental intervention. By incorporating professional psychological insights, the system can continuously improve, offering better support and achieving more reliable results in protecting adolescent mental health.

This is an optional funding source.

1. Blanchard LT, Gurka MJ, Blackman JA. Emotional, developmental, and behavioral health of American children and their families: a report from the 2003 National Survey of Children's Health. Pediatrics. 2006;117(6). Available from: https://doi.org/10.1542/peds.2005-2606
2. Reid MJ, Landesman Ramey S, Burchinal M. Dialogues with children about their families. Wiley-Blackwell. 1990;5-28. Available from: https://doi.org/10.1002/cd.23219904803
3. Nur Furqani ZA. The role of emotional intelligence in adolescent development. Atlantis Press; 2020. Available from: https://doi.org/10.2991/assehr.k.200120.058.
4. Schwarz SW. Adolescent mental health in the United States. 2009. Available from: https://academiccommons.columbia.edu/download/fedora_content/download/ac:126273/CONTENT/text_878.pdf.
5. Gleason MM, Goldson ED, Yogman MW. Addressing early childhood emotional and behavioral problems. Pediatrics. 2016;38(6). Available from: https://doi.org/10.1542/peds.2016-3025
6. Setiawati D, Endrastuty JF. Emotional maturity of vocational school students. Atlantis Press; 2019. Available from: https://doi.org/10.2991/icei-19.2019.83.
7. Sulistiowati NMD, Keliat BA, Wardani IY, Aldam SFS, Triana R, Florensa MVA. Comprehending mental health in Indonesian adolescents through mental, emotional, and social well-being. Compr Child Adolesc Nurs. 2019;42(sup1):277-83. Available from: https://doi.org/10.1080/24694193.2019.1594460
8. McLaughlin KA, Garrad MC, Somerville LH. What develops during emotional development? A component process approach to identifying sources of psychopathology risk in adolescence. Dialogues Clin Neurosci. 2015;17(4):403-410. Available from: https://doi.org/10.31887/dcns.2015.17.4/kmclaughlin
9. Young KS, Sandman C, Craske MG. Positive and negative emotion regulation in adolescence: links to anxiety and depression. Brain Sci. 2019;9(4):76. Available from: https://doi.org/10.3390/brainsci9040076.
10. Kumar M, Mishra R. Emotional maturity and academic achievement among adolescent students: A review of studies. Int J Indian Psychol. 2016;3(4). Available from: https://ijip.in/articles/emotional-maturity-and-academic-achievement-among-adolescent-students-a-review-of-studies/.
11. Cramer H, de Juan P, Tetreault J. Sender-intended functions of emojis in US messaging. In: Proceedings of the 18th International Conference on Human-Computer Interaction with Mobile Devices and Services. MobileHCI '16. 2016; 504-519. Available from: https://doi.org/10.1145/2935334.2935370.
12. Chitra V, Swaranalatha R. Application of emoji on social media: its impact among youngsters. Shanlax Int J Arts Sci Hum. 2021;9(2):17–20. Available from: https://doi.org/10.34293/sijash.v9i2.4193.
13. Aldunate N, Gonzalez-Ibáñez R. An integrated review of emoticons in computer-mediated communication. Front Psychol. 2016;7:2061. Available from: https://doi.org/10.3389/fpsyg.2016.02061
14. Li H, Su J, Chen Y, Li Q, Zhang Z. SheetCopilot: Bringing software productivity to the next level through large language models. 2023. Available from: https://doi.org/10.48550/arXiv.2305.19308.
15. La Rue EM, Li Y, Karimi HA, Mitchell AM. A description of the development and architecture of an SMS-based system for dealing with depression. Procedia Technol. 2012;5:670–8. Available from: https://doi.org/10.1016/j.protcy.2012.09.074
16. Majumder AKM, Veilleux CB. A smart cyber-human system to support mental well-being through social engagement. In: Proceedings of the 2020 IEEE 44th Annual Computers, Software, and Applications Conference (COMPSAC). 2020;134. Available from: http://dx.doi.org/10.1109/COMPSAC48688.2020.0-134
17. Coppersmith G, Hilland C, Frieder O, Leary R. Scalable mental health analysis in the clinical whitespace via natural language processing. In: Proceedings of the 2017 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI); 2017. Available from: https://doi.org/10.1109/BHI.2017.7897288
18. Dhuheir M, Albaseer A, Baccour E, Erbad A, Abdallah M, Hamdi M. Emotion recognition for healthcare surveillance systems using neural networks: a survey. 2021. Available from: https://doi.org/10.48550/arXiv.2107.05989.
19. Kusal S, Patil S, Choudrie J, Kotecha K, Vora D, Pappas I. A review on text-based emotion detection. 2022. Available from: https://arxiv.org/pdf/2205.03235v1.pdf.
20. Chen Y, Zhou B, Zhang W, Gong W, Sun G. Sentiment analysis based on deep learning and its application in screening for perinatal depression. In: Proceedings of the 2018 IEEE Third International Conference on Data Science in Cyberspace (DSC); 2018. Available from: https://doi.org/10.1109/DSC.2018.00073
21. Deng D, Zhou Y, Pi J, Shi B. Multimodal utterance-level affect analysis using visual, audio, and text features. arXiv preprint arXiv:1805.00625. 2018. Available from: https://arxiv.org/abs/1805.00625.
22. Balahur A, Turchi M. Multilingual sentiment analysis using machine translation. In: Proceedings of the 3rd Workshop on Computational Approaches to Subjectivity and Sentiment Analysis; 2012;52–60. Available from: https://www.researchgate.net/publication/262215708_Multilingual_sentiment_analysis_using_machine_translation.
23. Gupta S, Singh A, Ranjan J. Sentiment analysis: usage of text and emoji for expressing sentiments. In: Advances in Data and Information Sciences: Proceedings of ICDIS 2019. Springer. 2020;477-486. Available from: http://dx.doi.org/10.1007/978-981-15-0694-9_45
24. Shiha M, Ayvaz S. The effects of emoji in sentiment analysis. Int J Comput Electr Eng. 2017;9(1):360–369. http://www.ijcee.org/vol9/943-T048.pdf
25. Chen Z, Cao Y, Lu X, Mei Q, Liu X, et al. Sentimoji: an emoji-powered learning approach for sentiment analysis in software engineering. In: Proceedings of the 2019 27th ACM Joint Meeting on European Software Engineering Conference and Symposium on the Foundations of Software Engineering. 2019;841–52. Available from: https://doi.org/10.48550/arXiv.1907.02202
26. Ganie AG. Presence of informal language, such as emoticons, hashtags, and slang, impact the performance of sentiment analysis models on social media text? 2023. Available from: https://arxiv.org/abs/2301.12303.
27. Wolohan JT, Misato Hiraga, Atreyee Mukherjee, Zeeshan Ali Sayyed, Matthew Millard. Detecting linguistic traces of depression in topic-restricted text: attending to self-stigmatized depression with NLP. Proceedings of the First International Workshop on Language Cognition and Computational Models. 2018;11-21. Available from: https://www.aclweb.org/anthology/W18-4102.