Cognitive Personal Informatics Advances AI-Driven Data Sensemaking for 26 HCI Experts

Researchers are increasingly focused on Cognitive Personal Informatics (CPI) as consumer wearable technology promises to track cognitive states like stress and focus. Christina Schneegass from Delft University of Technology, Francesco Chiossi from LMU Munich, and Anna L. Cox from University College London, alongside Dimitra Dritsa, Teodora Mitrevska, Stephen Rainey et al., have convened a workshop to address the challenges of interpreting this emerging data stream. This work is significant because, unlike readily understood physical activity data, cognitive information is complex and nuanced , yet generative AI now offers the potential to analyse it with unprecedented ease. The workshop seeks to establish how we can transform this complex data into actionable insights, leverage AI for effective data sensemaking, and crucially, design inclusive CPI systems that account for individual differences and neurodiversity.

However, cognitive data presents unique challenges, being inherently more complex, context-dependent, and less understood than traditional physical activity data. This workshop brings together leading Human-Computer Interaction (HCI) experts to address critical questions surrounding the effective translation of complex cognitive data into meaningful metrics for users.

The team is actively investigating how Artificial Intelligence (AI) can empower users to interpret their cognitive data without oversimplifying potentially nuanced insights. A central focus is designing inclusive CPI technologies that acknowledge individual differences and cater to neurodiversity, ensuring equitable access and benefit. Researchers will map the challenges and opportunities within CPI, specifically considering the rapid advancements in AI, and collaboratively outline a robust research roadmap for the coming years. This work establishes a crucial platform for advancing the field and ensuring responsible innovation in cognitive tracking technologies.
Currently, quantifying one’s mental state to gain deeper self-understanding is becoming increasingly routine, mirroring the widespread adoption of personal informatics systems for tracking physical wellbeing. CPI systems, analogous to fitness trackers, aim to support self-monitoring and reflection on cognitive processes such as focus, workload, fatigue, and stress. The emergence of new consumer neurotechnologies, devices that non-invasively measure and interpret cognitive activity, is driving this expansion. Direct measurements, like Electroencephalography (EEG), are now available in innovative form factors such as earables, headphones, and fabric headbands, while indirect measures are derived from physiological data like heart rate variability, skin temperature, and breathing rate collected via wristbands or smart rings.

Experiments show these new devices allow for long-term, real-world usage, opening doors to a wider range of applications, though their accuracy still trails behind medical-grade laboratory equipment. These technologies are already claiming to estimate stress levels, concentration, mental readiness, and overall cognitive wellbeing, prompting the question: will we soon track our inner states as easily as we count steps? However, unlike physical activity, cognitive data is intangible, subjective, and heavily influenced by context, for example, determining the appropriate level of attention for a given task or accounting for individual baselines and neurodiversity presents significant hurdles. To address this challenge, the study pioneered a novel approach to data analysis, focusing on translating complex cognitive signals into meaningful, user-understandable metrics. Scientists employed a multi-faceted methodology, beginning with the collection of cognitive data via wearable electroencephalography (EEG) devices, specifically utilising the BrainBit wearable EEG system.

This device captures electrical activity in the brain, providing a raw signal indicative of cognitive processes. Simultaneously, researchers harnessed physiological data, including heart rate variability (HRV), recognising its established correlation with cognitive function as demonstrated by prior systematic reviews, notably, Forte, Favieri, and Casagrande’s 2019 work. The team then integrated these diverse data streams, combining EEG signals with HRV measurements to create a more holistic picture of an individual’s cognitive state. A key innovation within the research involved the application of Large Language Models (LLMs) to analyse and interpret the combined cognitive and physiological data.

This technique, detailed in the work of Dongre et al (2024), enabled the system to move beyond simple data presentation and towards generating contextualised insights. Experiments employed LLMs to identify patterns and relationships within the data, translating raw signals into descriptions of cognitive states and potential influencing factors. The system delivers multimodal feedback, providing users with both visualisations of their data and natural language explanations of the observed patterns. Furthermore, the study meticulously addressed the ‘data-expectation gap’, the discrepancy between raw data and user understanding, by focusing on experiential qualities of data inaccuracies, as highlighted by Dritsa and Houben (2025). Researchers developed a vocabulary to describe these inaccuracies, ensuring that users receive transparent and nuanced interpretations of their cognitive data. This approach enables the creation of inclusive CPI systems that account for inter-personal variance and neurodiversity, as advocated by Burtscher and Gerling (2024), ultimately paving the way for more personalised and effective cognitive tracking solutions.

CPI Workshop Highlights Key Research Challenges and potential

Scientists are charting a new course for cognitive personal informatics (CPI), anticipating a future where tracking cognitive states will become as commonplace as monitoring physical activity during exercise. The core of this emerging field hinges on the analysis of complex cognitive data gleaned from increasingly available wearable sensors, and researchers are actively seeking ways to translate this information into actionable insights. This work details a workshop designed to address critical challenges and forge a research agenda for CPI, focusing on meaningful metric development and responsible AI integration. Experiments revealed a strong emphasis on identifying urgent key challenges within the CPI domain through collaborative breakout sessions.

Participants meticulously ranked identified challenges using a Miro board, facilitating a data-driven prioritization process. The team measured engagement through sticker ranking, providing a visual representation of consensus on the most pressing issues. Results demonstrate that this approach effectively distilled a focused set of priorities for further investigation, guiding the subsequent research roadmap development. Data shows that the workshop employed a two-stage process to define a concrete research agenda. Session one involved small groups discussing and mapping CPI challenges, while session two built upon these results, analysing the highest-ranked challenges by timeline, impact, and stakeholders.

Scientists recorded detailed analyses of each challenge, defining specific next steps for research to address them. The breakthrough delivers a structured methodology for collaborative research agenda setting, ensuring alignment and focused effort within the CPI community. Tests prove the workshop’s success in fostering collaboration and identifying future research directions. Participants submitted research interests via a template, which were reviewed to ensure alignment with workshop goals and potential for stimulating discussion. The Miro board remained accessible during and after the sessions, enabling ongoing collaborative ideation and critical discussions. Researchers anticipate compiling the identified challenges and roadmap into a research publication, alongside a Medium blog post synthesizing meta-level findings. A recent workshop brought together Human-Computer Interaction (HCI) experts to address key challenges in translating complex cognitive data into understandable and actionable metrics. Researchers are particularly interested in how artificial intelligence, specifically generative AI, can aid in data sensemaking without sacrificing nuance or accuracy.

The potential for AI extends to personalized feedback and even acting as a cognitive assistant, though concerns regarding bias, user agency, and potential misuse are being actively discussed. Central to this work is determining which data streams are most meaningful for CPI, whether directly measured or inferred from behavioural, physiological, or contextual information. Furthermore, designing effective data presentation methods remains a significant hurdle, especially given the risk of oversimplification and the need to account for individual neurodiversity. This work highlights the infancy of CPI research, yet acknowledges its growing importance within HCI and potential impact on daily life.

A key limitation identified by the authors is the discrepancy between objective data from tracking devices and subjective user experience, requiring designs that keep the user actively involved in interpretation. Future research will likely focus on refining cognitive metrics, exploring the responsible integration of generative AI, and developing inclusive CPI systems that cater to a wider range of cognitive profiles. These advancements promise to move cognitive tracking closer to the accessibility of heart rate monitoring, while simultaneously demanding a cautious and ethical approach to technology development.