Getting started with personal science

From Personal Science Wiki
Jump to navigation Jump to search
Topic Infobox Question-icon.png
Linked pages on this wiki Tools (0),

Projects (0),

People (0)

This page is about how to get started with your own personal science project. See Getting started with wiki syntax for details on how to use this wiki.

Getting started with personal science can feel intimidating as there are a lot of components to consider when trying to answer personal questions. This page tries to accumulate best practices and advice to help both newcomers and experienced personal scientists in thinking and designing their personal science efforts. It is based on years of experience of personal scientists and is a living guide, which means that you are invited to edit and improve this article as well.

A framework for Personal Science[edit | edit source]

The original framework of personal science with 5 stages.

In 2020, Gary Wolf and Martijn De Groot published an article presenting a conceptual framework for personal science[1], which consists of five stages: "Questioning", "Designing", "Observing", "Reasoning" and "Discovering". Since then, this framework has been simplified in some places. Both the book "Personal Science" as well as the Quantified Self "Get Started" guide[2] have removed the "Designing" stage from their advice.

This simplification was done as it can be very hard to differentiate between "design" and "observation" as they often relate to each other and changes in the project design are often related to initial observations.

For most personal science projects this flow through the stages is not a singular occurrence but rather an iterative cycle that moves through these stages over time, as such it is fully expected that some initial reasoning after a short period of making observation leads to additional or changed questions which lead to slightly different or more observations over time and so on.

Questioning[edit | edit source]

The step of questioning requires thinking about why one wants to do a self-research project in the first place. It is also one of the most important steps in deciding to do a project, as it will inform all other decisions (e.g. on which methods or tools to use, which things to track etc.). There can be many reasons to start a self-research projects. Some common reasons that have motivated people in the past include,

  • Increasing self-awareness (e.g. when/where is particular thing happening to me)
  • Learning about the frequency or intensity of things (e.g. of symptoms like pain or allergies or changes in mood, weight or other phenomena)
  • Improving ones health or well-being, e.g. when having a chronic condition
  • Making progress in physical training or sports, trying to recover from an injury
  • Creatively expressing oneself through data and visualizations
  • Wanting to learn a new skill

Also see a more detailed exploration of reasons for and against self-tracking and quantifying yourself and different communities that approach self-tracking from different angles.

Framing a question[edit | edit source]

A good starting point for a personal science project can be to just write down a small bit on the goals and questions one is trying to answer with this. There is no "right" way for where or how to write this. One can write a few sentences in a physical notebook, somewhere in a digital notes tool or even create a small project page in this wiki (it can be as little as in this example). One doesn't have to share this brainstorming step with others, but by making it publicly available and discussing it with others one gets the opportunity to get feedback and advice early on.

In the context of health-related questions, it might also be better to start from exploring ones own experiences rather than trying to dive into trying to find causes right from the start. Collecting data to conclusively "proof" a certain diagnosis is a very high bar to take and can be an overwhelming task to start out in a personal science project. Instead, it might be worth focusing on questions like "How bad are my symptoms?", "How often do they occur?", "What helps to improve my condition?". Framing questions this way might make it easier to make observations and reason about the records one collects.

Observing[edit | edit source]

Once one has decided on what kind of questions one would like to pursue, it is time to think about how to answer it. Virtually all self-research projects require making deliberate observations. This requires selecting some parameters that one wants to pay special attention to to keep track of them.

What to observe[edit | edit source]

There are some things to consider when selecting what to observe.

  • Relevancy: Does observing this offer insights into what I really care about?
  • Convenience: Can I collect these observations easily and consistently?
  • Trustworthiness: Am I confident in the measurements?

On some level these things are interrelated. For example, using a digital tool such as a wearable device can be a very convenient way to passively collect data such as daily step counts. But is measuring steps really relevant for the question at hand? And how confident can one be in that the step counts are correct? In many cases it might be necessary to decide on trade-offs between these points.

Another good rule of thumb for observations is that the events or phenomena that one plans to record should be noticeable, variable and distinct as this will help facilitate recording the observations and reasoning about them later on. Having noticeable events helps recording the data, in particular when doing so manually by entering data by hand into a notebook, spreadsheet or mobile app (e.g. symptom recordings). Variability is important, as events that never change in frequency or intensity or any other dimension will not provide any insights (see relevancy above). Lastly, having distinct events is important to be able to correctly record observations as a lack of distinctiveness makes it hard to "correctly" count how often a thing happens.

Collecting dates & times[edit | edit source]

For virtually all types of observation one wants to record, it is worth to note down when this observation was made. Depending on the type of data one might only want to record the date of the observation or also the specific time and date of the observation. Most data entry apps and wearables might take care of this automatically, one needs to do it manually when using a notebook or spreadsheet. Unfortunately, "correctly" recording dates & times is not trivial, in particular when one moves or travels across time zones, but also due to Daylight Savings Time. See Dates & Times for more details.

Using proxies[edit | edit source]

Many things one might want to observe can be either hard to quantify (e.g. the abstract idea of "fitness") or hard to consistently observe because it would be too challenging to collect the "ideal" data over a period time. In such cases one can try to replace the direct observation through a proxy measurement that is closer to the ideal conditions outlined above. For example, instead of measuring "fitness" one could track physiological values such as Heart Rate Variability or how fast one runs, how many weights lifted etc.

Using self-assessments[edit | edit source]

Depending on the phenomenon one wants to observe, the use of self-assessments ("Rating of how I slept tonight", "Rating of my mood", etc) can be a viable option. Selecting the right range for this scale can be important. We frequently encounter 5-point scales (as in Amazon reviews) or 10-point scales (as in IMDb movie ratings). While broadening such a scale to give more nuance might be tempting, it also leads to the problem of differentiating between these values: "Should this be a 4 or a 5?". This paradox of choice can make it harder to do the self-assessment, leading to confusion, a potential abandonment of doing the observations or questioning the trustworthiness of the self-assessed observations.

In Personal Science, a recommendation is to stick to only three levels on a scale, for example for low/medium/high or positive/neutral/negative, to avoid such problems.

How to observe and record observations[edit | edit source]

There is no fixed "best" way to record any type of observations. Methods for recording data can be as simple as using a notebook in which one can enter all observations by hand, over using spreadsheet software like Excel or Google Sheets to the automatic recording of observations through devices like Wearables. More complex ways of recording data are not necessarily better than simpler approaches. Indeed, if a data recording protocol is too complex, it is more likely that one will not be able to do it consistently (see convenience above). One can also use a tool like Google Form to create a bespoke survey that one can fill out in regular intervals.

If one decides to use a wearable device or a specific mobile application for recording observations, it is important to check whether one can export and access all observations. If all data is stuck within a device or the data store of a manufacturer it might be impossible to use this data for a personal science project. If the tool in question has a page in the tools category one might find the answer there.

When to record observations[edit | edit source]

Unless one is using an automated tool that continuously records observations, one will have to decide when do record observations. There are two useful approaches for when to record observations. One can either record data every time one encounters a trigger or in relation to a routine prompt.

Trigger[edit | edit source]

In this approach one can record an observation every time something specific happens that comes to ones awareness. Ideally, such triggers are related to the phenomenon you want to observe and collecting all triggering events become the actual observations. E.g. if a project is about allergies one could record every time one sneezes. Or if a project is about is about intrusive pain levels, one can record every time experiences such pain.

Routine observations[edit | edit source]

If the events one wants to observe can not easily be related to triggers or are not as intrusive, it can be hard to use trigger approach. Instead the use of routine observations can be a viable alternative. While many self-research projects have played with automated alarms, experience shows that they are not very effective in the long term. While automated alarms as reminders can work in the short-term to provide reminders to record observations, their effectiveness quickly wears off due to alarm fatigue, resulting in not recording observations regularly.

To avoid this problem, linking manual observations to other routines can be a good way to regularly record data. Most people have fixed morning or evening routines. When a project requires only a daily observation it can be easy to add the recording of observations to these routines. Similarly, adding the recording to meal times can provide regularity for entering data. If more frequent recordings of observations are needed it could also be tied in into more routines which happen more frequently throughout the day if they are related to the topic of the self-research project.

Reasoning[edit | edit source]

Basically soon as one starts to record observations, one can start to reason about what one is seeing. There are myriads of ways to reason about data to gain insights from them, ranging from very simple to highly complex statistical methods or visualizations. This diversity of methods can make this step particularly daunting, but generally the majority of personal science projects that lead to interesting and meaningful insights do so using a range of simple methods to reason around the observations. This section gives some broad advice on things to consider when reasoning on personal science observations

Establishing a baseline[edit | edit source]

An example of a timeline visualization by Bastian Greshake Tzovaras. Red bars denote an external intervention that led to a change in observations. Observations have been summed up into daily totals for each category.

After establishing a personal science question and finding out how to collect observations it can be tempting to dive right into trying out any interventions ("let me drink less caffeine to see if my sleep improves", "How does diet X improve my weight loss", …). Before doing so it is advisable to collect enough data at a baseline before changing anything, as this will allow to compare the impact of any interventions against the "normal" state without any changes. Unless one has already recorded observations over a period of time to establish such a baseline, the start of a new personal science project is a good time to do so.

How much observations are needed depends on the question one tries to answer and the frequency of the phenomenon in question. If one looks at events that happen quite frequently only a week or two of observations can be enough for a baseline, but when dealing with rarer events it might be necessary to record a longer baseline. If one has a lot of historical data available, it might also be necessary to not use all of it as a baseline but rather limit oneself to more recent observations, as the baseline might otherwise reflect historic trends, life changes (e.g. changes in job, location, …) and interventions which might bias it.

Using a timeline[edit | edit source]

A surprisingly effective way to reason about ones observations is looking at if and how they change over time as it can highlight patterns. While there are many ways to visualize timeline data, for a start it might even be enough to just look at the observations one has recorded while they are in a spreadsheet, as in many cases this can already provide insights and facilitate reasoning about the observations.

If one decides to visualize the observations on a timeline there can be different strategies, depending on the granularity of the observations. If one recorded precise date & time information of events one can plot the data as it is or try to create bins that are easier to facilitate. For example, instead of visualizing each event on the timeline, one could sum them up into daily total values or averages. One does not need to be limited to creating daily bins, depending on the project and observations weekly or even monthly aggregates can be useful too.

An example of a binning visualization by Bastian Greshake Tzovaras. Shown are the number of tweets send, binned by the day of the week, showing a drop of tweets during weekends.

Timelines can also be useful when a personal science project includes doing an intervention and if one recorded data as a baseline. A timeline visualization can show if observations change after starting an intervention and if observations return back to the baseline after an intervention is stopped (see the first example visualization).

Binning observations[edit | edit source]

Binning observations can be an effective way outside the use in timelines. Many phenomena are influenced not only by interventions but by rhythmic changes, either because of behaviour changes (e.g. the difference between weekdays and weekends, working hours etc.) or external factors (e.g. seasonal effects in relation to allergies). To investigate this it can be useful to bin observations according to such factors (see second example visualization).

Discovering[edit | edit source]

In the personal science framework, the discovery stage focuses both on taking action following ones reasoning and on how to share the insights one has gained from doing a personal science project.

Considerations when sharing insights[edit | edit source]

Personal science projects take a first-person-stance by definition. Which means they are designed to answer a question for an individual or in some cases a small group of people (e.g. "Does this work for me? and "Does this work for us?"). As such, it is important to avoid the urge to make generalized claims that claim to work for everyone. These individual reports are valuable and useful.

Furthermore, regardless of what a project is about, an honest account of what one did is important. Things that one tried and which did not fully work out are just as informative for others as the success stories.

Where and how to share[edit | edit source]

Show and Tell[edit | edit source]

A common way of sharing insights from a self-research project within the personal science community are talks that follow the "Show & Tell" format, in which one presents a project along three main questions: 1. What did you do? 2. How did you do it? 3. What did you learn?

This Wiki has a rich database of self-research projects that were presented at Quantified Self conferences or meetups over the years within the category Show and Tell that can serve as inspiration.

Keating Memorial[edit | edit source]

Together, Open Humans and Quantified Self are organizing regular self-research meetings which happen every week and hold an annual online presentation event called the Keating Memorial. These weekly meetings provide a place to both share ones own self-research as well as discover what other people are doing research on.

Sharing a project on the Wiki[edit | edit source]

One can create a project page on this wiki within the Projects category. These pages can be free-form and can be used throughout the different stages of a personal science project, from the initial brainstorming to final reasoning steps.

Getting help[edit | edit source]

This wiki collects knowledge on self-research topics, tools used for personal science, as well as personal scientists and their projects, with the goal of supporting people in developing and implementing their own self-research. Additionally, there are weekly self-research chats that take place each Thursday at 10am Pacific / 7pm Central European time, which provide a way to share self-research projects and get feedback and help.

The forum of Quantified Self is another place where one can ask for advice and help[3].

References[edit | edit source]

  1. Wolf GI and De Groot M (2020) A Conceptual Framework for Personal Science. Front. Comput. Sci. 2:21. doi: 10.3389/fcomp.2020.00021