I spent about half my life in very rural areas of America learning how to hunt and fish and otherwise have a good time in the outdoors. For the other half, of course, I’ve been in cities of various sizes learnin’ fancy lawyer words and then trying to make some money with them.
I had some success along each of these paths. Which pretty much makes me too much of a redneck to fit in well in big cities while simultaneously making me too much of an over-educated liberal to really feel at home out in the country.
That problem aside, my somewhat unique cultural perspective can also be an asset. Like when I notice weird things happening with the autocorrect features on various applications on my Apple iPhone.
I was texting my cousin this morning using Facebook Messenger. I mentioned “snaring rabbits.” Despite having spelled that perfectly the first time, Facebook Messenger’s autocorrect feature changed my carefully selected phrase to “sharing rabbits,” which totally makes more sense I guess, what with Easter coming up and all.
This was not an isolated incident. I have a friend who lives in Missouri. I text him from time to time to ask about his level of success in “gigging” frogs (the best definition of a “gig” that I’ve found is “a harpoonlike device,” so when you are gigging frogs, you are spearing them, to make delicious frog legs). My iPhone sure likes to automatically change that to “gagging” frogs.
There is certainly a discussion to be had about humane methods of what is legally referred to in many jurisdictions as animal “take.” I just don’t know why my iPhone seems intent on manual strangulation as opposed to a quick and neat spear thrust.
I did not readily find academic research about artificial intelligence applications potentially being biased against more rural terminology (there is a whole bunch of research out there about AI being racist, however). Anecdotally, I have noticed this repeatedly, and it kind of makes sense logically based on how text-based AI features function.
What an autocorrect feature (or a large language model asked to write something) is doing, after having read through all sorts of content similar to what is currently being generated, is predicting what the human writer is most likely to want to say. Only about 20% of Americans live in rural areas. That makes for a lot more people texting, messaging, and otherwise writing about things going on in cities, suburbs, and medium-sized communities. AIs are training disproportionately on content created by people living in higher-population areas.
For the record, I did ask the latest version of ChatGPT, “Do you know anything about gigging frogs?” It totally nailed the correct answer. It also asked me why I was I interested in gigging frogs, so I explained that I was trying to ascertain whether it was familiar with terms more commonly used in rural America. It assured me it was.
While I was at it, I actually got kind of wrapped up in a whole conversation with the thing about this entire article. ChatGPT more or less agreed with me about how less-sophisticated AI used in autocorrect features works and thus how it often fails to correctly recognize niche terminology that may be more prevalent in rural America as it tries to serve the broadest group of users possible. The whole thing was equal parts impressive and uncanny.
Maybe all we’ve learned here is that the AI in use to correct our grammar in iPhones and on Facebook is quite primitive in comparison to OpenAI’s perhaps appropriately hyped platform. But if you ask me (or ChatGPT), texting and messaging applications currently in use certainly seem to be biased against language, like niche hunting and fishing terms, used primarily among smaller groups of people.
There are obviously greater injustices in the world. I can tell you, though, that messaging applications correcting my already correct text messages about rural things is an almost weekly annoyance, and one that makes me feel just a little bit more resentful toward big tech companies every time it happens. That can’t be a good thing for America’s rural-urban political divide.
Jonathan Wolf is a civil litigator and author of Your Debt-Free JD (affiliate link). He has taught legal writing, written for a wide variety of publications, and made it both his business and his pleasure to be financially and scientifically literate. Any views he expresses are probably pure gold, but are nonetheless solely his own and should not be attributed to any organization with which he is affiliated. He wouldn’t want to share the credit anyway. He can be reached at jon_wolf@hotmail.com.
The post Autocorrect, Other AI Applications, Are Biased Against Rural Language Like Hunting And Fishing Terms appeared first on Above the Law.
I spent about half my life in very rural areas of America learning how to hunt and fish and otherwise have a good time in the outdoors. For the other half, of course, I’ve been in cities of various sizes learnin’ fancy lawyer words and then trying to make some money with them.
I had some success along each of these paths. Which pretty much makes me too much of a redneck to fit in well in big cities while simultaneously making me too much of an over-educated liberal to really feel at home out in the country.
That problem aside, my somewhat unique cultural perspective can also be an asset. Like when I notice weird things happening with the autocorrect features on various applications on my Apple iPhone.
I was texting my cousin this morning using Facebook Messenger. I mentioned “snaring rabbits.” Despite having spelled that perfectly the first time, Facebook Messenger’s autocorrect feature changed my carefully selected phrase to “sharing rabbits,” which totally makes more sense I guess, what with Easter coming up and all.
This was not an isolated incident. I have a friend who lives in Missouri. I text him from time to time to ask about his level of success in “gigging” frogs (the best definition of a “gig” that I’ve found is “a harpoonlike device,” so when you are gigging frogs, you are spearing them, to make delicious frog legs). My iPhone sure likes to automatically change that to “gagging” frogs.
There is certainly a discussion to be had about humane methods of what is legally referred to in many jurisdictions as animal “take.” I just don’t know why my iPhone seems intent on manual strangulation as opposed to a quick and neat spear thrust.
I did not readily find academic research about artificial intelligence applications potentially being biased against more rural terminology (there is a whole bunch of research out there about AI being racist, however). Anecdotally, I have noticed this repeatedly, and it kind of makes sense logically based on how text-based AI features function.
What an autocorrect feature (or a large language model asked to write something) is doing, after having read through all sorts of content similar to what is currently being generated, is predicting what the human writer is most likely to want to say. Only about 20% of Americans live in rural areas. That makes for a lot more people texting, messaging, and otherwise writing about things going on in cities, suburbs, and medium-sized communities. AIs are training disproportionately on content created by people living in higher-population areas.
For the record, I did ask the latest version of ChatGPT, “Do you know anything about gigging frogs?” It totally nailed the correct answer. It also asked me why I was I interested in gigging frogs, so I explained that I was trying to ascertain whether it was familiar with terms more commonly used in rural America. It assured me it was.
While I was at it, I actually got kind of wrapped up in a whole conversation with the thing about this entire article. ChatGPT more or less agreed with me about how less-sophisticated AI used in autocorrect features works and thus how it often fails to correctly recognize niche terminology that may be more prevalent in rural America as it tries to serve the broadest group of users possible. The whole thing was equal parts impressive and uncanny.
Maybe all we’ve learned here is that the AI in use to correct our grammar in iPhones and on Facebook is quite primitive in comparison to OpenAI’s perhaps appropriately hyped platform. But if you ask me (or ChatGPT), texting and messaging applications currently in use certainly seem to be biased against language, like niche hunting and fishing terms, used primarily among smaller groups of people.
There are obviously greater injustices in the world. I can tell you, though, that messaging applications correcting my already correct text messages about rural things is an almost weekly annoyance, and one that makes me feel just a little bit more resentful toward big tech companies every time it happens. That can’t be a good thing for America’s rural-urban political divide.
Jonathan Wolf is a civil litigator and author of Your Debt-Free JD (affiliate link). He has taught legal writing, written for a wide variety of publications, and made it both his business and his pleasure to be financially and scientifically literate. Any views he expresses are probably pure gold, but are nonetheless solely his own and should not be attributed to any organization with which he is affiliated. He wouldn’t want to share the credit anyway. He can be reached at [email protected].
