The newly published data descriptor paper, LifeSnaps, a 4-month multi-modal dataset capturing unobtrusive snapshots of our lives in the wild, introduces a new public dataset empowering future research in different disciplines from diverse perspectives. Pervasive self-tracking devices have penetrated numerous aspects of our lives, from physical and mental health monitoring to fitness and entertainment. Nevertheless, limited data exist on the association between in-the-wild large-scale physical activity patterns, sleep, stress, and overall health, and behavioral and psychological patterns due to challenges in collecting and releasing such datasets, including waning user engagement or privacy considerations. The LifeSnaps dataset is a multi-modal, time, and space-distributed dataset containing a plethora of data collected unobtrusively for more than 4 months by 71 participants. LifeSnaps contains more than 35 different data types totaling more than 71M rows of data. The participants contributed their data through validated self-reported surveys, ecological momentary assessments (EMAs), and a Fitbit Sense smartwatch and consented to make these data available to empower future research. We envision that releasing this large-scale dataset of multi-modal data will open novel research opportunities and potential applications in multiple disciplines.
- Privacy and anonymity in accordance with the EU’s General Data Protection Regulation (GDPR)
- Large-scale data collected in-the-wild, where participants continued their normal daily routines
- Use of diverse human-centric data modalities with rich data granularity
- Emerging data types rarely studied until now, such as temperature, oxygen saturation, heart rate variability, automatically assessed stress, and sleep phases
- Community code sharing for reproducibility facilitating future research
A sneak peek behind the scenes
Innovative Training Network “Real-time Analytics for the Internet of Sports” (RAIS) aspires to provide 14 Early Stage Researchers a world-class training within a broad spectrum of subjects establishing a fertile interdisciplinary research and innovation community that will advance wearable technology, Block-chain Powered IoT, and Real-time Edge Analytics. The RAIS consortium consists of five partner universities in Europe (KTH, University of Cyprus, University of Insubria, FORTH, AUTH) and seven institutions and companies all around the world.
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement Innovative Training Networks (ITN) - RAIS No 813162
The RAIS Consortium Experiment officially started on May 25, 2021.
But the real work started way before, around nine months earlier…
Baffled by the lack of open datasets as early-stage researchers, we had the crazy idea of collecting our own, undaunted by the challenges of the feat.
Working within the RAIS Consortium, we collected data requirements from all our RAIS collaborators working on diverse research fields, from artificial intelligence to human-computer interaction to privacy and security. Based on these multifaceted requirements we drafted our study protocol and evaluated different technology products for suitability.
The questions abandoned: Which wearable device covers most of our data requirements? Do they provide sufficient documentation? Are there available APIs? Which option allows us to recruit from a larger participant pool? The answers to these questions led us to Fitbit Sense, Fitbit’s flagship smartwatch at the time, which had been released just a few months ago.
But before moving forward with any irrevocable actions, and in accordance with the requirements for ethical research and the GDPR regulation, we submitted our research protocol and consent form to the university’s institutional review board (IRB) for review and drafted a data management plan in collaboration with the university’s data protection officer (DPO), to do right by our users. On February 23rd, 2021 -just four months before the official start of our study-, we received the final green light, and we were delighted to kickstart our project!
Our potential participant pools were located in our partner universities in four different countries, namely Greece, Cyprus, Italy, and Sweden. Yet, physical meetings were impossible at the time due to the first wave of the COVID-19 pandemic and the strict lockdowns it introduced. That’s why, after the purchase of our Fitbit Sense devices we had to distribute them across our partners via registered post.
Each partner was then responsible for disseminating the call for volunteers and recruiting participants for the RAIS Consortium Experiment. The process was overseen by fellows at the Aristotle University of Thessaloniki and special care was given to a gender-balanced recruitment. In total, across both study rounds, we recruited 25 participants in Greece, 24 in Sweden, 12 in Cyprus, and 10 in Italy.
Throughout the duration of the study, all participants received weekly hand-drafted emails (with useful Fitbit tips!) encouraging them to continue contributing to the study and asking them to complete certain surveys.
We also adopted many monitoring processes to have a first look at statistics on compliance in real-time to encourage less engaged participants to contribute to the study.
Once the study’s nine weeks had passed, we had to, unfortunately, bid our participants goodbye. We asked them to complete their last surveys and share their data with us, again reminding them of the importance of data privacy and security for the RAIS Consortium and the scientific community. And we did hold true to our promise; a dedicated data anonymization team was working for months after the end of the user studies to verify the anonymity of the LifeSnaps dataset before publication.
A real-world use case
An indicative real-world use case of LifeSnaps is the mental healthcare sector. Finding out the trajectories of a patient's psychological traits often requires repeated verbal interactions. Traditional in-person interviews are not always preferable because of the economic burden on clinicians and patients. Additionally, self-reported questionnaires can be problematic as they are based on patient recollections and self-representation. These two tools can be combined and enriched with passively and objectively collected behavioral data. That's exactly what LifeSnaps is, and hopefully, it will facilitate research in the mental health domain, advancing prevention and treatment.