Lifelogging #

What? #

I define lifelogging as the practice of keeping track of the events in your life within some kind of physical medium, like a computer’s hard drive. The “Quantified Self” concept is closely related to this, although I see it as being more specific about what it is focusing on (namely numerical measurements).

Why? #

There are two primary reasons why lifelogging interests me:

1. Lifelogging promises to provide a perfect record that I could refer back to when trying to remember something. Some examples of things that would be nice to remember more accurately:
   • The ideas in an engaging conversation
   • The places I was in the last day where I should look for my missing keys
   • The stuff I did during the last week when trying to figure out what projects to continue working on.
   • The stuff I did over the last weekend when someone asks “how was your weekend?”.
2. Finding correlations in, or otherwise analyzing, lifelogging data could lead to clear insights about which behaviors lead to better health, more desirable states of mind, and better life outcomes however you want to define them.
   • See my Tracking Health page for more details.

An Ideal System #

The primary issue with lifelogging is that collecting even small amounts of information consistently can be very time consuming. This is exacerbated if you are not running a specific experiment on yourself that requires very targeted collection, but are rather trying to paint a picture of your life that could be used for more generic analysis and memory-aiding.

Therefore, my ideal lifelogging system would be totally automatic. This could be achieved with the following components:

• A GPS monitor logging my location.
  • Done using GPSLogger
• Tracking food consumption via photos.
  • Done using autojournal + Google Photos.
  • Plan to move to MomentoDB for this.
• A wearable camera that sees what I see and stores the video somewhere.
  • Partially done using a snapcam
• A wearable microphone that hears what I hear, and transcribes all heard speech to a searchable text log, like a script to the play that is my life.
  • Not done yet, see https://roberdam.com/en/wisper.html (https://news.ycombinator.com/item?id=33608437) for prior art!
• Programs that log what I’m doing on my various devices (laptop, phone, etc.).
  • Almost done with ActivityWatch.
  • Previously/currently done with selfspy.
• Tracking Exercise
  • Should be inferrable from my location tracking, but I haven’t written the logic yet.
• Tracking habit adherence. Like exercise or practicing a skill.
  • I found proactively assigning myself to do a habit at a specific time works very well, and this has the added benefit of tracking when I tick off my assignment as “done”. I use Google Tasks for this, which show up on my Google Calendar. Ideally I would write code to save my task data so that it could be visualized in other places (and so deleted tasks wouldn’t disappear from history).
• Various sensors that track mood and health metrics.
  • Manually done with MomentoDB, therefore not consistent :/
  • Previously done with Nomie, which IMO is strictly worse than MomentoDB.
• Logic that tries to infer meaningful “events” from the datastreams collected by these components. For example, image processing software analyzing my camera feed coupled with GPS data could infer that I was playing soccer from 1-2pm on Saturday.
  • Not done yet
• A place to store all the collected data.
  • Done with Google Drive
• A calendar or timeline -like visualization system that provides a scrollable, searchable way to browse all the generated life events.
  • Partially done with autojournal.
  • See continuous glucose monitoring experiment for a timeline view.
• A tool that can find correlations in the stored data and in general condense it down to insights. Preferably this would be folded into the calendar or timeline visualization.
  • Partially done with biomarker correlator

An example of what this looks like can be found at https://miguelrochefort.com/blog/calendar/ (not my system exactly, but very similar).

Currently Available Tools #

This table summarizes the functionalities of different software tools I’ve tried for personal data tracking. Here are what the columns mean:

Data Entry/Export: Descriptions of how possible/easy it is to enter data into the tool and get it back out for custom storage or analysis. They use this legend:

• 🚫 = Impossible
• ⛏️ = Manual and Difficult
• 🛠️ = Manual and Ergonomic
• 🤖 = Automatic

Analysis: The tool provides some way to analyzer and visualize your data to see trends over time or other views into the data.

DB Comparison: The tool compares your data to external data sources as part of the visualization to help put said data into context (e.g. is my value too high or too low).

Recommendations: The tool makes specific recommendations about how you should behave given your input data.

Each column value follows this legend:

• 🚫 = Feature Doesn’t Exist
• ✅ = Feature Available

*ActivityWatch automatic export is in progress.

General Methods #

I try to store as much personal data in Google Drive as possible. I find it convenient to access both manually and programmatically. I’ve written a program called “autojournal” to parse data from my Drive (and some other places) and display it both in graphs (see my CGM page) and on my Google calendar.

For tracing of general information, I’ve found Momentodb for Android to be super useful. It can track almost any data type with very few taps, AND it syncs to Google Drive (Sheets) automatically!

Categories: Lifelogging

Backlinks: Biomarker Correlator, Continuous Glucose Monitoring, Tracking Health, Camera, Daily Habits, Task Tracking,