 Imagine how a pianist’s fingers glide across the keys, each note reflecting their emotions and state of mind. Now, consider how your fingers interact with your phone’s keyboard every day. With texting becoming our primary form of communication, the keyboards on our phones have become as central to our lives as a piano is to a pianist. At BiAffect, we believe these everyday interactions hold the key to understanding our mental health. With BiAffect, we aim to provide unique insights into mental health through the analysis of typing dynamics on smartphones. Similar to what how a Fitbit helps track physical fitness, BiAffect tracks mental and emotional health—like a Fitbit for the brain. Our mission is to offer a non-intrusive, continuous method of monitoring mood and cognitive changes by analyzing how people interact with their keyboards. With the increasing preference for texting over talking, our phones have become essential to our daily communication. Many people spend significant time typing out messages, emails, posts, and other text rather than engaging in voice conversations or phone calls. This shift in communication style means that our keyboards are now central to how we express ourselves and connect with others. By analyzing our typing patterns, we can uncover valuable insights into our mental and emotional states, making it possible to track our well-being in a meaningful way. BiAffect began life as the prizewinning entry to the Mood Challenge for ResearchKit, and the team is led by Alex Leow, MD, PhD, Professor in Psychiatry, Biomedical Engineering, and Computer Science at the University of Illinois at Chicago, with ongoing technology development spearheaded by senior AI/ML/MLOps expert Theja Tulabandhula, PhD, AI researcher and software engineer Andrew Paparella, senior software engineer Shannon Young, and research specialist Faraz Hussain. BiAffect has been covered extensively in the news over the years, including in the Wall Street Journal, Rolling Stone, and IEEE Spectrum. Leveraging Keystroke Dynamics for Mental Health Insights BiAffect is based on the premise that typing patterns—how we interact with our keyboards—can reveal a wealth of information about our mental and emotional states. By continuously analyzing keystroke dynamics, we can detect subtle changes that may indicate mood shifts or cognitive fluctuations. Core Keystroke Dynamics Features Typing Speed Typing speed, measured by examining the statistical distribution of flight times, can reveal changes in processing speed, a key domain of cognition. Variations in typing speed can indicate different mental states. For example, a decrease in typing speed may be associated with depression or cognitive fatigue, while an increase could suggest heightened alertness or anxiety. Flight Time Flight time is the interval between consecutive key presses. Longer flight times can indicate hesitation or uncertainty, which may be linked to inattention, distraction, or cognitive overload. Shorter flight times can suggest increased urgency, anxiety, or impulsivity. Error Rates Error rates refer to the frequency of corrections and typos. Higher error rates can be indicative of cognitive stress, distractions, etc., and are commonly observed during severe depression. While individual error rates may vary, significant changes over time can provide additional context when analyzed alongside other keystroke dynamics. Key Hold Time Key hold time, or the duration a key is pressed, can provide insights into motor function and cognitive control. Prolonged key hold times may suggest motor impairments or cognitive slowing, while shorter hold times can indicate increased motor activity or restlessness. Accelerometer Data In addition to keystroke dynamics, BiAffect also incorporates accelerometer data from the user’s smartphone. This data helps capture physical movements and phone orientation information that can provide important context about the kinds of activities the user was engaged in, such as whether they were walking, standing, laying down, in a moving vehicle, etc. while typing on their phone.  Machine Learning Analysis The extracted features from keystroke dynamics and accelerometer data are fed into our machine learning models. We use a combination of supervised and unsupervised learning techniques to identify patterns in the data. Supervised learning models are trained on labeled datasets, where the mood or cognitive state of the user is known. Unsupervised learning models, on the other hand, help uncover hidden structures in the data, providing insights into previously unknown patterns.  Real-Time Monitoring and Feedback BiAffect provides a dashboard within the app where users can monitor their typing patterns and accelerometer data. This dashboard presents the analyzed data, helping users understand their mental health trends. By examining changes in their keystroke dynamics and active task scores on activities such as a Go/No-Go Reaction Time test and Trailmaking test, users can gain valuable insights into their mental state and take proactive steps to manage their well-being.
Conclusion At BiAffect, we are committed to advancing mental health monitoring through innovative technology. By leveraging various keystroke dynamics such as typing speed, flight time, error rates, key hold time, along with accelerometer data and active tasks, we provide valuable insights into mental health. Our custom keyboard plays a crucial role in this mission, integrating seamlessly into our app to capture and analyze these metrics with precision and efficiency. By continuing to refine our algorithms and expand our feature set, we aim to make BiAffect an indispensable resource for mental health management. We welcome you to download BiAffect and see how our pioneering use of keystroke dynamics can enhance your understanding of mental health. Our aim is to provide practical insights that can help you navigate your well-being.
0 Comments
 Longtime BiAffect collaborator Dr. Michelle Chen chaired a symposium at the 2023 meeting of the International Neuropsychology Society in Taipei, Taiwan entitled “Smartphones as a window into everyday brain health: applications of keystroke dynamics, ecological momentary assessment, and accelerometry” featuring BiAffect Principal Investigator Dr. Alex Leow as discussant alongside fellow presenters Dr. Tammy Chung and Andrea Cladek.
 BiAffect Principal Investigator Dr. Alex Leow chaired a symposium at the 25th Annual Conference of the International Society for Bipolar Disorders in Chicago, IL on “Digital Phenotyping to Support Wellness in People With Bipolar Disorder” featuring fellow presenters Drs. Maria Faurholt-Jepsen and Debbie Huang.
BiAffect team member Claudia Vesel has been awarded a travel grant to present her research on "Diurnal Patterns as Evidenced by over Eleven Million Smartphone Keystrokes During Daily Usage" at the 2019 NNDC Annual Conference, to be held September 24-25 in Ann Arbor, MI.
 On the 30th birthday of the World Wide Web, BiAffect's Dr. Alex Leow speaks at Chicago AI Days about the use of artificial intelligence in measuring mood.
BiAffect chosen by the NIH to present at the 2018 MD2K mHealthHUB Technology Showcase on June 4, 2018. Drs. Ajilore and Leow are excited to demonstrate BiAffect as an open science research tool to improve the reliability and validity of passive sensing using keyboard metadata.
|
