[FloorEgg]

Hey there is a company examind.io that has a product that lets you assign writing assignments for students to do on their own time in their UI that runs in the browser, and it tracks every keystroke and interaction by the student and analyzes it to determine whether it's the student writing or if they are copy/pasting or transcribing it from AI. Also it has option to give student access to AI research assistant right in the software so you can inspect how the student is using AI to help with their work.

I think it might give you the assurance you need while giving the best experience and opportunity for your students to do honest / hight integrity work with the latest (AI) tools.

The benefit of forcing students to write their grades work in person is you know it's coming from them, but the downside is it's a very artificial test not representative or what real world work will be like. I think examind can give you the best of both.

[em-bee]

i can use AI to generate text and then type it manually into that tool.

[FloorEgg]

It evaluates the process the student used to write the essay, not only the final result. It makes it transparent to the professor. So the professor will see the student typed it out word for word. this is in contrast to an authentic essay writing process that involves a lot of editing.

[viraptor]

There's also going to be that one kid who knows what they're doing and write a fake typer app for others. (I was that kind of kid...)

[oerdier]

I don't feel assured at all. I don't want to bet on any horse in the volatile arms race between AI and anti-AI.

[FloorEgg]

Re-reading my original reply and your response and I think we had a misunderstanding. I never intended to make you feel assured with my post. I was trying to communicate that the features the product provides could help you feel assured that the student actually completed the work themselves and if they used AI to help, you can see exactly how. (And that they appropriately paraphrased, etc)

The whole point of the product is to give professors more flexibility in the kind of assignments they use (and even allowing students to use LLMs in a controlled way and be evaluated in how they use them) while ensuring academic integrity.

For example allowing students to use LLMs as research assistants and even to help them consider and structure ideas, while ensuring the student paraphrases everything sufficiently to prove they actually understand it and can put it in their own words.

To be clear I understand and respect your desire to protect the integrity of the diplomas and credentials you are giving out (especially in contrast to the so many who let cheating run rampatnt), but at some point you may want to be able to accurately evaluate how students use industry-standard tools. (like when calculators were first introduced).

So sure, be skeptical, but maybe be careful about throwing the baby out with the bathwater.

[oerdier]

I appreciate your taking the time to elaborate. I see you've responded to all comments on your original comment and remained civil despite people's negative sentiment. There are too few places on the internet left where such civility remains, and I thank you for contributing.

I would not feel assured of students actually completing the work themselves with _only_ something like examind.io as an extra measure. For it to be used in their own time, they would use it on their own hardware. As user viraptor pointed out, whenever there's anti-cheat software, someone is going to create targeted anti-anti-cheat software. That's what I meant by arms race.

For me to feel assured of students not fooling the anti-cheat software, for their input device they would have to use hardware controlled by me. It's not feasible to let them use hardware controlled by me in their own time.

I can see how a tool such as examind.io might help in accurately evaluating how students use other tools on a computer. For that they could use hardware controlled by me, during a test.

[FloorEgg]

Their approach is to bring maximum transparency into the process the student used to write the essay, rather than the final result.

I don't really see how it's about an AI vs anti-ai arms race.

It's not my company and I'm not responsible for selling it so I'm probably doing a poor job...

But if you want to evaluate your students writing and ensure integrity and also provide them with longer windows to work on bigger writing assignments (and even allow them to use LLMs to help them write in accordance with your rules) then wouldn't an application like this help you?

I don't understand why I'm getting such a negative reaction from everyone for sharing this... I genuinely thought I was helping by pointing out a solution to your problem...

[joatmon-snoo]

Assuming you're the founder, this is the type of BS comment that makes the rest of us hate AI founders.

It's vacuous, makes vague claims that don't leave room for proof/disproof, and doesn't offer any reason that it's any better than a prompt that asks GPT4o "was this generated by AI y/n"

[FloorEgg]

I am not the founder. Also I pointed out they track user actions and keystrokes and analyze them, which is clearly distinct from just pasting the students work into an LLM and asking if it was generated by AI. I'll go further to say that they can tell if the student left the browser window and for how long. Also natural essay writing patterns are different than transcribing something from another source. I'm pretty sure they have more methods but I don't remember them all.

Your comment is full of unnecessary animosity and resentment and I think it is inappropriate.

I know the founders and I also know they have very happy customers.

Just because you have experienced some vapor ware startups or whatever doesn't mean every company is one...

I shared it with the OP because it sounds like they care about the problem the product solves.

Lastly, I want you remind you that the very first line in the community guidelines is "Be kind. Don't be snarky". I find your comment both unkind and snarky.

https://news.ycombinator.com/newsguidelines.html