Creating the “Trends” Page on This Site

This is the trends page

I created an interactive visual representation of the connection between my mood and the things that I take pictures of.

I've been uploading images with tags since March 2020 and have been tracking my mood since August 2022. It seems obvious to connect the two, but I just hadn't up until this point. Today I did it.

First – why does this matter?

This project – this page, this website – is less about me as an individual and what makes me "unique" but more about me as a part of a faceless mass of people who consume technology. The function of any consumer technology — Android, Macbook, SmartTV, React, Claude (yes, dev tools are consumer technology.) — today is to surveil the masses and provide data on us to the ruling class (Peter Thiel, personally). This (and more) is what they see, this is why the technology exists. Identity today is best defined by a statistical composite of responses to stimuli that are developed by large technology companies. This project is me "spilling my guts", flipping my digital skin inside out, making my algorithm bare.

The idea

Most days when I wake up and get on the computer, I log my mood and upload pictures from the previous day. Doing it first thing in the morning makes it something of a constant in this scientific experiment. It makes it so that the excitement or mundane-ness of the previous day doesn't affect my mood and rather I'm tracking a general vibe.

Some days I don't think to upload images for a couple days, and sometimes I upload pictures and don't update my mood. But in general I upload images and log my mood. The pictures that I upload are all pictures that I take. The images that I take are usually taken to document and remember things.

I'm more likely to take a picture of something very mundane than I am to take a picture of something very exciting, because when something exciting is happening I am experiencing it and usually don't think to take a picture.

So, there is a match between my images and my mood tracking – they are both sort of "averages" of my human experience. If I mush these two data sources together, I should be able to find some rough approximation of how stimulus and activity makes me feel. I expect to find:

• connections I wouldn't have made otherwise
• false positives
• uncomfortable truths

How it's Built

I have a database that I can query via an api to get data representing my tags and mood. On the backend I am already querying every image, tag, and mood prior to implementing this feature, so I pass all of that to a function where I map each day to my mood. Then I go through all the tags and I find all the days with moods that are present on days that that image has been uploaded. I group all of those together, and then I calculate the average mood on days when I've uploaded an image that contains a given tag. I compare this against the average mood on ALL days and subtract this value to get a delta value. I do basically the same thing with my colors, but I don't want to talk about that because I need to rework that, though I suspect if the quality of that gets better then I will have something there that's meaningful.

When my website (which uses SSR to generate HTML via React app) builds, I inject this data into the page, and have chart.js build the graph of this in a <canvas>. One snag I ran into was learning that mobile browsers do not render canvas elements that are too tall, probably to preserve memory / battery. This thing was like 40,000 pixels tall (at the time of writing I have ~560 unique tags, and each bar is 80px tall.) I fixed this issue by virtualizing the list.

I didn't want results like one data point on one day that was really good. There is a counterpoint that having an outlier mood score AND an outlier tag could be causal, but I'm betting against that – showing this will result in a lot of data that's meaningless. I institute a rule that for a tag to show up in this list the tag has to appear on at least three separate days.

More on frequency of upload: the bars are color coded going from light -> dark based on how many distinct days a tag has been uploaded on. More distinct days provides more data and a more accurate signal of the impact of given stimuli on my mood.

Results

When I was building this I had an idea of what it would look like but my data in my development environment is stale (although it's real) so I didn't really know. It was very rewarding to push to prod and see myself. Some interesting results:

• travel is the highest scoring tag with a good amount of data. Although it's not really a tag I think to use that often, I've done a lot of traveling and there are only 7 distinct days. I like traveling, it makes me happy and excited. This is a signal that this model contains truth.
• molly is my ex who I have mixed feelings about now but I definitely had a lot of fun with her. She scores highly on this. I met her in Chicago and we moved to LA together. My memory of living with her includes making dinner together and getting yelled at. I was really happy while I was dating her. The relationship had its ups and downs but she was an incredibly supportive partner when she wasn't losing it at me, and we did a lot of fun things together and went on a bunch of sick vacations. A lot of LA-adjacent tags score very highly. Ktown, wilshire, etc.
• yves was an old roommate who I still think is really cool but she created a very chaotic living situation for me. There are only three pictures of her, all three of which represent good memories in my mind, but she scores the lowest of any tag – by far. Did not expect that. Obviously this is based on limited data, but is this a function of underlying feeling?
• original cartoons and cartoons score very poorly. I guess I don't like staying inside and drawing. I do it a lot but I don't think I spend all day drawing very often.. if ever? Sometimes I like to stay up late drawing.
• skateboarding, camping, and truck all score very high. I like gettin dirty! Although truck is mostly pictures of other peoples' trucks. Maybe I just like being outside. Trucks are mundane, they're everywhere. Maybe when I'm in a mood where whatever's on the corner excites me enough to take a picture of it, I'm happier.