Students figured out their tests were graded by AI and the easy way to cheat | Hacker News Reader

Students figured out their tests were graded by AI and the easy way to cheat

theverge.com

65 pointsuptown26 comments

Love it. If their future is half-baked code judging them, gaming the system is a necessary life skill! Grading homework is an NLU approximation to self-driving cars, we’re nowhere near.

If I was a parent of one of those kids, I’d be really angry at the school system to have this kind of setup.

I find this really sad. I do congratulate the students that managed to figure out the system and game them (they will go far!), but using this keyword salad grading thing is totally missing the point of education.

I came from a third world country where the "education system" is basically just a series of tests on memorization and conformance, and I'm hoping I would never see a system like that ever again.

'I came from a third world country where the "education system" is basically just a series of tests on memorization and conformance...'

2x2 is 4... sorry third world you say?

The correct answer is "2+2 is 4"

Your score: 0/1

When you’re given a philosophy-related task, whether it’d be reading and reviewing or a dissertation… the professor expects your response to be exactly xyz or else you’re penalized.

So yeah, third world :).

Calling it "AI" seems generous. It's really just searching for keywords. If it was any decent AI then it could detect incoherent jumbles of words.

Calling it "AI" brings in the clicks

These school teachers are doing less and less work and demanding more and more money -- even when schools are closed and there are bound to be some savings to be had just on HVAC, maintenance, supplies, etc.

I'm also amazed at the cushy deal this "AI" company must have with the school district, selling them software that just looks for a few keywords for, most likely, very lucrative terms.

> I'm also amazed at the cushy deal this "AI" company must have with the school district, selling them software that just looks for a few keywords for, most likely, very lucrative terms.

What industry do you work in that doesn't have garbage software licensed for top-dollar prices?

they are demanding more money because they are paid badly to begin with.

https://www.thebalancecareers.com/average-teacher-salary-by-...

In history class this is what was required on certain questions of my tests. I could have a correct answer without a certain keyword and get a lower grade on that question because of it

Good and practical learning.

As well, to get a job you have to put a list of keywords which will fit well that one which the recruiter has.

That way, the AI to replace us one day doesn't have to advance toward us, we may just easy degenerate toward it (as to succeed you have to put accurate list of words which you don't have to understand and AI never had).

BTW, What do you call a blonde who dyed her hair black?

The combination of

a) simple keyword scanning

b) prompting for thought out sentences

is downright unethical, IMO.

We have these discussions here sometimes, but if you were ever involved in a project like that, I feel like I'd want to whistle blow, or more likely just quit.

The SAT essays have been graded by computer for years. They also have a human grader who glances at them for 2 minutes. When I learned the criteria they use it kinda exposes what a sham the whole thing is.

SATs are multiple choice, this is apples and oranges to the situation in the article.

The SAT added an essay section in 2005, increasing the maximum score to 2400 from 1600.

The third word in my comment was "essays".

Wow, the essays are computer-graded? Please tell me more.

SAT is not computer-graded, but by 2 humans, apparently.

The GRE is graded by a human and a computer, and a second human in case of diverging scores.

SAT: Responses to the optional SAT Essay are scored using a carefully designed process.

Two different people will read and score your essay. Each scorer awards 1–4 points for each dimension: reading, analysis, and writing. The two scores for each dimension are added. You’ll receive three scores for the SAT Essay—one for each dimension—ranging from 2–8 points. There is no composite SAT Essay score (the three scores are not added together) and there are no percentiles.

https://collegereadiness.collegeboard.org/sat/scores/underst...

GRE: For the Analytical Writing section, each essay receives a score from at least one trained rater, using a six-point holistic scale. In holistic scoring, raters are trained to assign scores on the basis of the overall quality of an essay in response to the assigned task. The essay is then scored by e-rater®, a computerized program developed by ETS that is capable of identifying essay features related to writing proficiency. If the human and the e-rater scores closely agree, the average of the two scores is used as the final score. If they disagree, a second human score is obtained, and the final score is the average of the two human scores.

https://www.ets.org/gre/revised_general/scores/how/

This is new to me. Interesting.

If the teachers don't grade anymore, what they are doing with all the free time ?

Driving Uber. Teaching is a labor of love, it doesn't pay. /s

And in the mean time prepare plans for lessons, answer questions, bringing resources for students use - they never were mostly or exclusively occupied by grading.

They’re not bringing resources - the classes are online.

I haven’t encountered this platform before, but most textbooks now come with class plans for teachers (especially at university where they are used to kill off second hand book sales)

> They’re not bringing resources - the classes are online.

I mean, digital resources, bringing them ready to be used in class. Files have to be found, or better created from others by cutting and coupling, they have to be named, put in correct place for accessing. Even though the material sometimes doesn't change much form a year to another, it needs to be checked and adjusted for current class trajectory.

Teaching is a hard intellectual work. Good teaching, that is.

Online classes aren't generally considered any easier to teach. And preparing them in the first place is often completely left to the teacher, who's entire workflow just got thrown out the window.

And the materials provided with the textbook often suck, and require more work to use than creating your own from scratch.

They are still probably holding class meetings online.

And the effort of converting everything to online in the first place can be pretty overwhelming.

I can actually see some utility in this.

First formulate a list of keywords, then stretch them into sentences, resulting in an information dense structure that hits all the important points.

Reply on news.ycombinator.com