Process, methods & tools
For my project, I selected personal informatics (quantified self) and found an open FitBit dataset.
It was a week-long research and design work grounded in a critical literature review.
Phase 1 – Literature review
I went through about 30 academic papers and books to learn more about Quantified Self (QS) in general and about current advancements in QS visualisations.
I soon realised that, although there were models describing the life cycle of personal informatics, I didn’t encounter a model of what visualisations should be provided to the user in various contexts and stages.
To consolidate all the insights from the literature review phase, I decided to carry out a thematic analysis. This led me to devise my own framework that I called a “Model of visualised self”.
Phase 2 – A model of visualised self
The model builds on the insights made in the reviewed papers and consists of seven stages - these are by no means clearly separated, however, they have distinct characteristics:
1. Exploration
Allow users to explore long-term time series data [10], it should be presented in the context of other QS data streams creating a coherent multi-faceted view [10,11] and avoiding data silos. Start with a simple view that can be gradually enhanced with more details on demand [3].
2. Discovery
Facilitate discovery by highlighting anomalies and discrepancies in the data [6]. Showing global trends, periodic patterns, and multiple visual representations of the same data enables new interrelationship discoveries [3,6]. User history, goals and contexts are all important factors that help seeing the data in new ways [11]. Make it easy to tag and annotate events to create a richer picture of the context surrounding an event – richer information can lead to uncovering hidden relationships [12].
3. Reflection
Support viewing results from different perspectives [1] by presenting surprising or puzzling facts – e.g. compare user performance with friends, family members or general public [1]. Bring past records into user’s attention unexpectedly to promote reflection [6]. If possible, model and show alternative outcomes of different goal-achieving scenarios [6]. Present goals and key metrics set by the user to help them reflect upon data. Allow insights to be shared with friends or public and facilitate feedback from them.
4. Diagnosis
After exploring the data and reflecting on discoveries, the users are ready to diagnose the causes of discrepancies. Allow them to annotate the data and store the discoveries and insights they have made. Assist them in turning these into actionable points [3].
5. Action
Support users in adjusting, setting and achieving their goals [3,11]. Suggest next steps. Gamification can motivate action – e.g. collecting rewards [16], taking care of pets [13] or plants [2,8] or use of avatars [7,15] were shown to be an effective approach to behaviour change facilitation. Involvement of family members has been shown to improve chances of a positive change [14].
6. Maintenance
Display current status and make data interpretable at a glance [3,11]. Offer a simplified view but allow for gradual enhancement [3]. Use positive reinforcement – users do not want to see bad results [10] – abstract representations of the data may be perceived better than charts [5] or living metaphors [2,8,13]. Take into account that users will forget or skip measurements and do not punish them for it [4].
7. Discrepancy
Support users by notifying them about discrepancies between their current status and their goals [11].
Phase 3 – Design focus & Rationale for proposed visualisations
I focused on the Exploration and Discovery phases described in the model above as these presented the biggest opportunity for a novel approach.
There seemed to be a gap in the literature around discovering periodicity in QS data despite the fact that young professional’s life is, in some ways, inherently periodic.
Spiral visualisations like the QS spiral [9,17] present an interesting alternative to traditional time-series charts. A continuous spiral timeline can allow users to discover repetitive or periodic patterns in their data.
Although looking promising, the QS spiral has not been tested with users and therefore its efficiency is not known. Judging from the provided pictures, it is not very easy and straightforward to navigate. Could there be a way to visualise data periodicity while making the visualisation easy to navigate?
Tools I used
- Pencil and paper
- Exploratory.io (a data science tool based on R)
- MS Excel
- Adobe Illustrator
