10
votes
TietoEVRY, a software company from Finland, has developed a new font called Polite Type which uses machine learning to rewrite offensive language into more inclusive forms
Link information
This data is scraped automatically and may be incorrect.
- Title
- Font automatically blurs swearing and changes slurs to stop cyberbullying
- Published
- Aug 21 2020
- Word count
- 322 words
So, I have a few problems with this.
First, it's not really a font's responsibility to handle this kind of logic, it just feels to me like the wrong layer to handle this at. You can filter text on the client or server when it's just a string, you don't have to complect the rendering with the filtering. Because of that complection, this only works with this font -- if you want to use a different font -- because of how it looks -- you completely lose this functionality.
Secondly, it's not implemented well: it fails the classic Scunthorpe test and instead returns "Svulvahorpe". It also renders "ho" as a blur, so I guess Christmas is cancelled if you use this font now?
A corollary to this point is that it is trivial to bypass this by replacing characters or adding extra spaces or full-stops to bypass the whitelist.
Finally there is a philosophical issue for me here. I can understand that not being exposed to the content of bullying messages might soften the blow somewhat, but there is still someone hurling abuse at you, and even if you can't see what they've said, you still know they've said something that's meant to hurt you. Surely the correct response is to just report it and delete the message? I hope the moderation team don't have to use this font or it will make their lives much harder.
Overall, seems kinda cool from a design/typography perspective, but from a technical and pragmatic perspective, poorly thought out.
Rationalwiki's section on the euphemism treadmill does a good job explaining this, as well as how using person first language solves the actual problem instead of attempting to make up new terminology: https://rationalwiki.org/wiki/Euphemism#Euphemism_treadmill
I found the Person-first language page on RationalWiki (linked to from the Euphemism page) was really low quality and didn’t do a good job at describing the concept from a neutral perspective.
It’s honestly fascinating how obsessive some people are about trying to use language in creative ways so as to try to minimize offensiveness, but it’s also disheartening to see so much effort spent on futile efforts like trying to walk these metaphorical treadmills expecting they’ll actually move forward within the metaphorical space.
Damn, yeah, that article is EDGY. I was just referring to the subsection under euphemisms, which is higher quality in spite of it's brevity and it not being the article's focus.
Yeah, the inconsistency of quality of rationalwiki is unfortunate. I thought the Euphemism page was pretty good, so thanks for linking that.
I agree that it's not correct to inject functionality like this at the font level. However, I also think this is very cool.
Fonts are one thing that users generally have control of, irrespective of their platform. That makes novel uses of fonts like this more appealing.
It doesn't have to be changing the meaning of words, either. It could instead offer brief dictionary definitions of complex words, or offer acronym resolution. Sure it'd be clunky and probably break a lot of things, but it's still a very interesting idea!
Here's the link to try out the font (it's pretty hidden in the article).
"Once there was a not traditionally beautiful barnacle. He was so not traditionally beautiful that everyone died. The end."
🤣
Oh boy... I wonder who got to code in all the slurs that were to be replaced.
Hmm...it kind of reminds me of the stand-up bit Key & Peele did before this sketch, where they describe that people who ascribe negative connotations to otherwise mundane words/things might have a problem, themselves. (Of course, they end the stand-up bit by saying the following sketch is not one of those instances (the way they said it was obviously way more funny)).
Anyhow, it's an interesting technical display and certainly there are plenty of words that do have inherent negative connotations, but the list is in constant debate and flux...so I don't know how they'd keep up. And the implementation is kinda clumsy. It'll flag a word if it contains a portion of a bad-word even before you complete the non-bad-word that you've meant to type. So..it's like almost forcing you to think of the bad word when you weren't even thinking of it.
Which, as far as I can tell, means they don't actually want this font to be used - not that it could be used effectively, its "censor" doesn't even recognize all-uppercase forms and they say as much - as much as it being a symbol, a novel way to get people talking about the issue.
It's going to upset all the usual suspects, but maybe the technical aspect will mitigate it for some - I'm unsure. I can't seem to find the implementation details, and it doesn't look like it is doing much beyond a
.replace(..., ...).Not ligatures, contextual alternates. They can do things like changing some glyph(s) based on the characters that surround it: see the
*placement in the linked example, and see Datalegreya for why I don't consider this to be particularly exciting.And if companies, as a group, were famous for pulling weird stunts to advertise their anti-bullying campaign - while activists and artists such as font designers notable for their boring, conservative projects - I might even believe it.