Many organizations have employee well-being programs, but even with these in place, mental health issues that need clinical intervention and/or counseling go undetected. Often help arrives only after the problem is very severe, or when it has resulted in adverse outcomes. Behavior, Business, and Social Sciences (BBSS) research at TCS supports proactive initiatives for associate well-being. We discuss here some elements that can enable realistic modeling of organizational agents and an easy to consume interface that can improve well-being among employees.
Download the full article or click the button to continue reading.
A 24-year-old man with chronic depression takes to drugs and alcohol unable to deal with stress at work; a 30-year-old woman ends her life because she is deep in debt due to multiple family responsibilities; a middle-aged telephone operator hangs himself in office in the middle of his work day because of multiple health issues. It is not uncommon to see such headlines in newspapers. To many of us, these may not appear to be strong enough reasons to hurt ourselves, or end life. What becomes apparent though, is a need for greater awareness of mental well-being and support for those with mental health issues at the workplace.
Constant and rapid change often requires today’s workforce to not only adapt to external unforeseen changes but also have sufficient psychological resources and capabilities to respond to change in an effective manner. Capabilities such as resilience, stress management, and sound decision-making are some of the often-cited employee-related strengths critical to an enterprise’s success as well. The TCS R&D division has well-being as one of its main focus areas in the Behavior, Business, and Social Sciences (BBSS) research area. At BBSS, we use a multidisciplinary lens to understand states of well-being within the enterprise and to guide associates towards improved well-being.
Need for Multi-disciplinary Approaches to Well-being
Many organizations have employee assistance programs, but even with these mechanisms in place, many problems and challenges that need urgent clinical and/or counseling attention go undetected or reach the employee assistance program only after the problem is very severe and has resulted in adverse outcomes such as self-harm or suicide. This is due to several factors, ranging from the stigma associated with mental health problems, privacy concerns, and the lack of mental health awareness, to name a few.
Given the sheer scale of the problem of mental health incidence worldwide, and the criticality of its management for the individual, enterprise, and society as a whole, there has been a sustained interest in the use of digital technologies to develop more scalable solutions for mental health management. Emergence of areas such as telepsychiatry,1 digital mental health,2 and digital therapeutics,3 reflects the growing application of technologies in the field of mental health. While there have been significant strides made in the development of these technologies for mental health, key concerns regarding robustness, scalability, and evidence-driven technologies continue to dominate the field.4
Digital Solutions for Well-being
Building a scalable digital solution which can help address issues such as early identification of burnout, stress detection, and management and prevention of self-harm at the workplace not only demands expertise in the behavioral sciences, but also in related areas (e.g. workplace norms, HR policies) and in areas such as image recognition, voice analysis, big data management, and analytics, among others. The solution has to have several components such as:
A knowledge base that supports behavior modeling related specifically to the workplace
Image and voice recognition algorithms that can detect anomalies in the subject
A platform that can fuse data from various sources and anchor the solution
An engaging mobile app that serves as a companion and confidant to the user
A robust measurement and intervention system for enhancing well-being in the enterprise
BBSS research has focused on each of these components and we would like to briefly elaborate on them here.
A Knowledge Repository
The need for repositories as sources of information in a domain has been a critical requirement across scientific disciplines. There are knowledge repositories related to behavioral sciences, but the focus here must be on:
A structured and usable repository amenable to computational behavioral modeling, synthesis of knowledge, generating evidence, and identifying and generating new areas for research.
The repository has to be a structured entity, such as a set of tables in a well-recognized standard entity such as a relational database management system (RDBMS). This is to facilitate both the entry/modification of information into the repository as well as its querying to obtain its contents. Note that querying and consumption of the content should be possible both by a human or a program.
The creation of a minable repository of a vast area such as human behavior requires many steps such as:
1. Creating a corpus of papers in the behavioral sciences domain
2. Identifying relevant papers using a low-cost multi-class classifier
3. Identifying sections in each relevant paper that have empirical results of interest using intra paper classifiers. (This second-level classifier is built because techniques such as information extraction in natural language processing are compute-intensive.) The process splits as below:
3.1 Information tagging and extraction techniques to identify behavioral variables, the relation terms, the strength and the confidence of the relation, as well as other information considered relevant from text, using a text relation miner;
3.2 A table extractor to process the tables (if any) in the paper to extract quantitative empirical findings
4. Combining the information obtained through the relation miner and table extractor to put together a complete record of an empirical finding using an integrator
5. Verification of the extracted knowledge by domain experts, after which validated knowledge is now moved to the main behavior repository
Several challenges have to be overcome in creating the repository: lack of domain experts as well as varied structures and formats of the reports and tables. There is also the issue of the lack of definitive solutions for processing tricky areas of natural language and written text. A separate machinery may be required to create resources for activities such as correcting error terms, mapping synonyms to each other, expanding acronyms and abbreviations, identifying other semantic relations between terms such as hypernymy and hyponymy, adding context to a term in a particular usage so that there is a differentiation between different uses of the same term, among others.
Modeling Realistic Organizational Agents
One of the goals of the development of the Behavioral Relations Repository is to demonstrate the use of hypotheses from behavioral and management sciences in modeling behavior across the organization. Agent-based modeling and simulation (ABMS) is amenable to the study of a wide range of behavioral dynamics in a given context. ABMS has been traditionally used in the study of complex systems, but its applications to organizational psychology have been sparse (Hughes et al., 2012). ABMS has been applied to the study of macro behavior in the realms of social psychology; however, its use in studying the micro-foundations of behavior has been limited. To bridge this gap, fine-grained agent-based models can be used to develop realistic models of behavior anchored in the domain of organizational behavior.
A Well-being App
There are many mobile apps that are meant to help with measuring and improving one’s well- being, but many of them are meant for those who are already seeking help or already know they have a problem. Today, it is possible to sense emotions through voice, facial expressions, and activity. A mobile phone today can sense so much about its owner: acts of physical activity, tone of voice, facial expressions, and lifestyle behaviors (sleeping/shopping/spending/interacting), among others, that it becomes a source of information about a person’s well-being. Within the limits of GDPR, much of this data can be analyzed and alerts may be provided to the subject/caregiver on anomalies from usage patterns within the wellbeing app. If the app is further gamified and endears itself to the user, both accurate data collection and intervention may be possible.
An Anchoring Platform
Central to a digital solution is a platform for multi-modal behavioral data capture through numerous data capture devices, sensors, and secondary data sources; behavioral knowledge mining from multiple sources; behavioral analytics; behavioral knowledge delivery to various simulators, games, and reasoners. The platform acts as the hub for the entire ecosystem of tools and applications. It must have the Behavior Knowledge Repository as its source of knowledge. The platform must have analytics capabilities such as a personal well-being dashboard, population level wellbeing analytics, and a pattern discovery module. Other capabilities include a manual intervention module and a behavior study design wizard in addition to the behavior relations repository and an interventions repository. In keeping with current GDPR policies, the platform must be compliant with privacy and other aspects outlined in standard GDPR practices that are in effect today.
The Road Ahead
In building the digital solution with the components mentioned above, we have had a taste of success (View Box). The knowledge repository that we are building will keep growing as behavioral science is a vast area encompassing many domains such as psychology, cognitive science, and sociology, among others, and new knowledge is added constantly across all areas of science. Challenges in building the knowledge repository include the reproducibility debate, which focuses attention on the need for replicating evidence to test for its robustness and generalizability across different settings.
Advancements in NLP and related areas have helped evolve novel approaches in addressing these challenges in creating such repositories, and work on these aspects is underway at BBSS. The serendipitous value we have discovered is that many other areas of research are keen on learning our techniques in building knowledge repositories in other domains as well.
Both the digital platform and the app will gain new features and usage through our continuing field trials. We envision these tools being used in enhancing employee well-being as well as other contexts where human behavior measurement and change is of importance. This could range from problems in the space of change management, competency management, learning, and many more.
Digital Companion for Well-being
TCS endeavors to be proactive in employee support initiatives. Being a technology company spearheading digital transformations for leading businesses, it uses digital tools for several employee initiatives such as fitness, marathon running, accessibility and social responsibility. As mental well-being is a top concern for TCS HR, it has approached the BBSS research team for various studies notably in workplace stress and productivity among support service teams; individual productivity and well-being; and the link between perceived supervisory support and subjective and objective performance measures, among others.
We were asked to build a scalable solution to foster mental well-being. The solution has come from our research in BBSS and is anchored on a platform called Beacon. It draws knowledge from the domain repository we built, it analyses facial expressions and voice anomalies of the user, it holds the mobile app we call ‘T’ (‘T’ is an internal identifier), and it has scoring mechanisms and intervention protocols. We would like to highlight the consumer end of the solution ‘T’ here.
‘T’ is a digital companion based on the concept of a pet dog . The 1990s saw the rise of virtual pets. Since a large percentage of TCSers are millennials, the idea of a virtual pet is an accepted concept. Our Serious Games research team created the character and tested it with the target audience who said virtual pets are fun to engage with, and valued the chance to share emotions without being judged.
‘T’uses the latest techniques in the area of purposeful games . It is data privacy-conscious and built upon numerous sensing technologies to measure well-being and assist people in adopting behavior changes that help them stay in the positive end of the well-being spectrum. We have carried out initial user pilots with ‘T’ and are currently refining ‘T’ to incorporate these inputs as well as carry out larger pilots across the enterprise. In terms of physical sensing, ‘T’s capabilities include emotion detection with speech, text, and camera images as inputs. Other capabilities include laughter and blow detection, physical activity sensing, and heart rate sensing.
Scenarios for ‘T’
‘T’ is currently envisioned as a Digital Companion App for well-being at work including measurements and interventions targeted at workplace stress management. Other behavioral change interventions that can be enabled by ‘T’ include a more balanced lifestyle, improved relationships, enhanced motivation, and other well-being behaviors of interest. In addition, ‘T’ is also amenable for applications in other scenarios such as managing depression. Figure 2 illustrates this application. As shown in Figure 2, ‘T’s sensing, measurement, and intervention capabilities enable an individual living with depression to track his/her moods, activities, and lifestyle over time.
As a companion for therapy for individuals with depression, ‘T’ presents a promising scenario that is being explored with our partners in healthcare and insurance contexts as well as with clinicians working with patients suffering from depression. Many patients require a combination of both medication and therapy to manage their symptoms and to stay productive. ‘T’s capabilities potentially allow for the therapeutic interventions to be delivered to the user at prescribed times as well as be sensed/tracked for medication and therapy follow-ups with the clinicians. Given that the doctor’s appointments are likely to be short and sporadic, ‘T’ allows the patient, clinician, and the caregiver to manage the treatment regimen longitudinally. Once we have a more robust version of ‘T’ ready for the workplace well-being context mentioned above, we plan to explore ‘T’s applications for depression management (Figure 2).