Is it cheating to use spell check?
By @edent
· AI MSc NaBloPoMo · 10 comments · 600 words · read ~1,449 times.
When I was a kid, our school had one computer per classroom. Luxury!
Teachers had long-since given up on the state of my handwriting. So I got special dispensation to write up some of my work on whatever primitive word processor was installed on the PC. With one caveat: no spell check!
Which, even as a ten year old, I thought was reasonable. Learning to spell is an adult life skill. So using a spell-checker was cheating.
Once I got to secondary school, it was assumed that I knew how to spell. And there was no restriction on using Wrd 2.0's splel chek.
But I had one teacher him hated an grammar check. He thought it encouraged laziness of writing. Which, to be fair, it probably did.
Of course, come exam season, there were no computers. All answers were laboriously hand-written. The use of spell- and grammar-checking software had left me lazy. And my handwriting had atrophied beyond all hope. It's a miracle I passed!
Spell Checking is Artificial Intelligence.
It doesn't seem like it because it is so ubiquitous now. But it is more than just a computer looking up words in a list. There's a huge corpus of learning which goes into spotting homophones, predicting which word you meant, and calculating the likeliest candidates based on your previous writing.
I'm now in the middle of writing up my MSc dissertation. I'm doing it in Google Docs - because I'm a masochist.
Google Docs has a useful feature called Smart Suggestions which → offers to autocomplete your sentences.
Is that cheating?
I've signed a declaration to my University saying that my dissertation is all my own work. But it isn't.
Google's AI suggested a couple of dozen sentences. Google's AI proofread and corrected both my spelling and grammar. Just how much of my work was generated by 2022's Clippy?
The general consensus from my teachers is that it is not cheating.
Students are now using the GPT-3 AI to write their entire essays.
Is that cheating?
I think most people will instinctively say yes. A whole essay written by machine is cheating.
An autocompleted sentence is fine. But a whole essay is not. Where's the line? To reinvigorate an old joke "We've already established that some AI is cheating - now we're just haggling over character count".
If I need a paragraph summarising the history of mayonnaise, do I need to waste my time writing that - or can I outsource it to an AI? Doing so doesn't prove I understand the material. But is it materially worse than taking someone else's work and rewording it into my own style?
Perhaps AI isn't cheating when it is merely scaffolding?
When Clippy offered to help with writing a letter, it usually meant that it would show you how to format an address and whether to end with "Yours faithfully" or "Yours sincerely". Is it cheating to use a well-structured template?
This ramble-post is to say that I don't know. My essays are mostly my work. When the AI has suggested something, I've been free to reject it. But I rarely do. Google Docs knows what I want to write before I write it. Am I merely the editor of machine generated text?
One paragraph of this blog post was generated by OpenAI. Can you spot it?
Does it matter?
@Edent one per classroom, wow. When I started secondary we had one for the whole school (CBM PET), and by the time I left we had three (BBC B and two BBC A iirc). Well, technically there were probably another three still, the original CBM PET and another and somewhere a Spectrum that I don't think ever got used. Shortly after I left one of the teachers grabbed a job lot of MSX for a computer room (HCC didn't want them so the school could get them cheap). Different times!
@Edent I'm glad that worked for you. It was the opposite for me. When doing English essays I used to draft on my Atari ST and then have to hand write them out as the school wouldn't except typewritten versions. I wonder what schools do now?
Carla N says:
"Except" typewritten versions? was that an intentional misuse to demonstrate the fact that one has to be smarter than IA? I can't count the number of times I have been "corrected" by spell-check when I was not wrong. You still have to know your (NOT "you're!") stuff.
@Edent I had to tell the staff how to disable spell check when I took my exams at school 😕 I often used to use some primitive wordprossecor thing and later a newton based device
The question about which paragraph was generated was tricky! I found myself scanning for paragraphs without a personal slant, which was only really the one starting "It doesn't seem like it because it is so ubiquitous..." But maybe that isn't a good indicator?
DinoNerd says:
My concern about spell check, grammar check, suggested responses, and for that matter entire computer-generated texts is that they aren't very good. I long ago established that both my spelling and my grammar were more correct than would be produced by the "artificial intelligence". I could use it to flag typos, producing a hybrid better than either - but at least a third of the suggested corrections would convert correct text to incorrect text, particularly with grammar.
In recent years, I've noticed people converging on what-the-computer-suggests - i.e. Microsoft, Apple, and Google software engineers seem to be defining correct written English for many people, not all of them ESL speakers. After reading tons of this artificially enhanced prose, many errors no longer look wrong to me. Obviously, language changes over time - but this particular mechanism does not appeal to me.
Completely computer generated text hits new lows - there's zero reason to expect it to be accurate, regardless of the quality of its English usage. Obviously plenty of humans create lies, fake news, exaggerations and propaganda. But at least with them there's likely to be a motive one can use to help detect dubious statements. The computer, on the other hand, does not understand what it's saying - it just puts words together in patterns that follow rules it's inferred from writing samples. Color me unimpressed, except perhaps with the potential for mischief.
@Edent Using spell check isn't cheating. Highlighting incorrect spelling is useful tech for people with disabilities. People with ADHD, dyslexia, vision impairments & other disabilities may have issues spelling words, even if they know how to spell them correctly. Sometimes people spell words differently for cultural reasons or if it's a second language.Is having a human edit your work also cheating? A blanket policy like that doesn't sit right with me
LDV says:
Spell check (and its sibling, grammar check) have the goal of bringing something that already exists but is not conforming to a shared standard (the spelling of a word, comma usage, etc...). I'm sure an argument could be made for variances in dialect and such, but I also think most people would agree that language works better when we are using the same rules.
A fair amount of the AI is creating things that do not exist yet. That includes links between ideas. Could you write some thoughts on The Tell-Tale Heart by Poe? Probably, if you've read it or are familiar with it. Could an AI write it for you even if you've never heard of it before now? Even if you only let it finish your sentences, you could probably get something mostly correct. And if you did know about it, an AI could point you to links you wouldn't have considered.
If you're being primarily tested on spelling, then spell check is cheating. Otherwise, it's just an automated way to do a bit of professional clean-up. I can't think of a way that using something to create or link ideas isn't cheating in at least a school setting.
Also, to reply to the final line: it does matter if I can tell what you added that was made by AI and what wasn't, at least to me. I don't follow blogs to get a mindless mash-up of aggregate ideas. I follow blogs like yours because I want to know what the human thinks. If I wanted an AI's opinions on a topic, I'd generate it myself.
You could make an argument that you're still curating what's posted by rejecting what you don't like. Fair enough - that's the same as a link blog. But link blogs clearly indicate that the content isn't created by the curator of the link blog but by someone else.
I apologize if this has come off as rude. I don't mean to be. But I cannot express strongly enough how bad I think it would be for untagged AI-generated content to become the norm. I don't want the "dead internet" theory to come true.
What an interesting short essay. I thought the reframing of the old joke to be “haggling over character count” was an excellent way of making the point. I’m not normally short of an opinion, but at the moment on this question I just don’t know!
Eddie Spatz says:
All of the opinions posted (above) would probably be tagged as plagiarism by scholastic standard bearers at any given higher education institute. There are probably many others who express these ideas in a simalar way. There is no accounting for general language usage Each student is expected to generate completely origional sentence structure, grammer, and spelling, etc. I don't know how an individual avoids being accused of plagiarism.
More comments on Mastodon.