People have been writing a bit about “estimated reading time”, the feature where a link to something is accompanied by an estimate of how long it will take you to read it.
Matt Campbell pointed out that estimated reading times are often ableist. Adrian Roselli pointed out that existing widgets such as scroll bars do a better job of giving you a sense of how long of a read a page is.
But, from my perspective as somebody who still regulary works on web-based reading systems, the biggest argument for why they might not be ableist is that they are largely bullshit. They aren’t that useful – at least not in the way people think.
Reading time of any given text is affected by a number of different variables:
- Word count or length.
- The writing style.
- Formatting and design.
- Typeface, different typefaces effect people’s reading speeds differently.
- Medium (screen vs print).
- Quality of the medium (bad screen/newsprint vs hidpi).
- Individual physical factors (astigmatism can make dark mode reading illegible, other physical conditions can make light mode illegible).
- Vocabulary (anything with a lot of unfamiliar words takes longer to read no matter the other variables).
- Skill at reading this particular kind of writing in this particular format (being able to read a book quickly does not necessarily transfer to being able to read something quickly on the phone, and vice versa).
- Hydration and other physical factors that affect focus.
To be meaningful, a reading time estimation algorithm needs to take a number of these variables into account. The bare minimum would combine text complexity, usually estimated as a combination of sentence complexity (writing style) and vocabulary, with length and the median level of schooling of the assumed audience to get a rough estimate. This would only even be vaguely correct for a reader with the estimated schooling.
So how do all of these sites and tools estimate reading time?
Well, they usually just divide the word count by 233 (or similar number) and rounding to the minute. It’s basically just a word count transposed into an idiosyncratic and opaque base 60. At some point a programmer read in a study that the average person read 233 words per minute and decided that this would be a great way to estimate reading time for everybody on the fucking planet. This average person does not exist any more than the half child from the average 3.5 person family does. It doesn’t take anything whatsoever into account.
The more “sophisticated” estimation algorithms add an arbitrary number of seconds to the reading time for every image, but because it can’t take the type or contents of the image into account, the number of seconds they add is effectively just a random number.
So, estimated reading time calculators are literally just a method for obscuring the word count.
What should you do instead?
Well, sites and communities that are tailored specifically to active readers, like Archive of Our Own, tend to just use word counts plus metadata that lets people estimate what sort of text it is. AoOO uses tags. A reader can look at a fan fiction listing and pretty accurately guess whether they can fit reading it into what remains of their lunch break
An objective measure, like word count, combined with metadata provides the reader with enough information to accurately estimate their own reading time
A more general approach, more suitable for link blogs so inclined, would be to use word count with a note on writing style. Something like “850 words, fluffy, no long words” or “850 words, tech jargon, complex sentences” would be much more useful than the “3 minutes” that most default today
Basically, estimated reading times are useless to the reader. Give them something useful instead.