Slashdot Mirror


How AI Can Spot Exam Cheats and Raise Standards (ft.com)

AI is being deployed by those who set and mark exams to reduce fraud -- which remains overall a small problem -- and to create far greater efficiencies in preparation and marking, and to help improve teaching and studying. From a report, which may be paywalled: From traditional paper-based exam and textbook producers such as Pearson, to digital-native companies such as Coursera, online tools and artificial intelligence are being developed to reduce costs and enhance learning. For years, multiple-choice tests have allowed scanners to score results without human intervention. Now technology is coming directly into the exam hall. Coursera has patented a system to take images of students and verify their identity against scanned documents. There are plagiarism detectors that can scan essay answers and search the web -- or the work of other students -- to identify copying. Webcams can monitor exam locations to spot malpractice. Even when students are working, they provide clues that can be used to clamp down on cheats. They leave electronic "fingerprints" such as keyboard pressure, speed and even writing style. Emily Glassberg Sands, Cousera's head of data science, says: "We can validate their keystroke signatures. It's difficult to prepare for someone hell-bent on cheating, but we are trying every way possible."

47 comments

  1. False Flag by Anonymous Coward · · Score: 1

    School: Student, you are expelled permanently because our AI says you cheated.
    Student: I didn't!
    School: You have no recourse, we keep your money, you are gone, your life is ruined, goodbye!
    Student: 'Tis a fair court!

    1. Re:False Flag by arth1 · · Score: 2

      Indeed. When a person's whole future life is on the scales, the weight needed can't just be "upon preponderance of evidence" or "beyond reasonable doubt". Even a 100,000:1 risk of a false positive, when combined with the data that around 3.8 million Americans will obtain a university degree this year, that means that dozens of them will trigger false positives and risk their degree taken away from them by machines and the machine-like humans who operate them.
      When faced with science that says it's only a 1 in 100,000 chance that you're innocent, good luck convincing enough review board members that you're an exceptional case and innocent.

    2. Re:False Flag by gweihir · · Score: 2

      Well, if I were on such a review board, I would likely find that the "educators" were so incompetent, that I cannot actually judge the performance of the student.

      --
      Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
    3. Re: False Flag by Anonymous Coward · · Score: 0

      Yeah, humans never fail of course. That's without entering into personal bias or pettiness.

      Just changing your example to a human judge, everything else exactly equal in the exchange, makes it fair?

      At least with technology we may have accountability.

    4. Re:False Flag by Anonymous Coward · · Score: 0

      Ha ha! Yes! Good one!

      The "educators" are universally incompetent. What a jape! What a jest! I see what you are driving at, good sir.

      Perhaps you in your lofty position, with your IQ of well over 190 could illuminate us all on just how we should go about running an education system.
      No doubt you would refer to that droll "Three 'R's" chestnut people like you drag out every time education is mentioned, or perhaps you would lament the use of some standard or another. "New Math" or the so-called "Core Curriculum" come to mind.

      Come sir. Tell us all your deep wisdom. We wait with baited breath!

      Incompetent Educators! It is to laugh.

      Asshole

    5. Re:False Flag by jellomizer · · Score: 1

      There is a general problem of simultaneously people who trust computers too much and not enough.
      The same person who would expel a student for cheating from an AI Algorithm report would be the same person who manually counts the number of rows in a database to make sure the computer has it right.

      AI tool are not fool proof, but they are good at reducing the workflow. Other then manually checking 1000 papers for cheaters, you will only have a dozen to review. Then you will need to use human judgement to see if the AI is correct or not.

      The goal of AI isn't to be perfect but good enough. Just as a school shouldn't rely on one person to determine if a student gets expelled it shouldn't rely on an AI.

      --
      If something is so important that you feel the need to post it on the internet... It probably isn't that important.
    6. Re: False Flag by arth1 · · Score: 1

      At least with technology we may have accountability.

      How, exactly? Can we send the machine to jail so it can ponder its mistakes, and so other machines will have less desire to do bad deeds?

    7. Re:False Flag by Anonymous Coward · · Score: 0

      I too prefer a German-inspired scorched-earth policy towards market competition. Always degrade college students and new graduates. It's the only way old gweihir can stay employed!

    8. Re: False Flag by Aighearach · · Score: 1

      It can be subjected to code audits.

  2. The false positives by Anonymous Coward · · Score: 0

    will make this a nightmare.

  3. AI by 110010001000 · · Score: 1, Insightful

    So are we just labeling every algorithm that detects patterns "AI" now?

    1. Re:AI by Anonymous Coward · · Score: 1

      So are we just labeling every algorithm that detects patterns "AI" now?

      Yes.

      Any other questions?

    2. Re:AI by 110010001000 · · Score: 2

      Yes. So I can call myself an AI expert now? Also, let me know if Eliza is now considered AI.

    3. Re:AI by Anonymous Coward · · Score: 0

      Why not? It worked for Paul Simon. link

    4. Re:AI by gweihir · · Score: 1

      You know, when the fraudsters feel they have to inject a minimal bit of honesty, they call it "weak AI", which is the AI without "I". The whole terminology is a big, fat, bald-faced lie. There is no AI at this time and it still is completely unclear whether it is even possible. No, the demented physicalist argument "but humans do it" does not count, as we have absolutely no clue how humans do it. At this time, even "magic" is a scientifically possible explanation. (What, you thought Science rules out magic? No, it does not. It just needs to be repeatable and measurable and "magic" becomes a Scientifically sound thing.) That tells you how extremely little we know.

      --
      Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
    5. Re:AI by 110010001000 · · Score: 1

      Unfortunately a lot of taxpayer money is being thrown at this boondoggle.

    6. Re:AI by Anonymous Coward · · Score: 0

      Yes. So I can call myself an AI expert now?

      Yes, you have my permission. ;-)

      Also, let me know if Eliza is now considered AI.

      I don't know, ask it maybe?

    7. Re:AI by angel'o'sphere · · Score: 1

      As I pointed out several times on /. ... it is not you who defines what AI is.
      Just pick a random university and check what they teach in their 'AI classes'.

      You make the common mistake mixing up SF movie 'thinking machines' with what scientists call AI.

      --
      Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
    8. Re:AI by gweihir · · Score: 1

      As I happen to be a scientist, it actually is me who defines what AI is. (Along with a few million other people though.) I am sure that goes way beyond what you can grasp, though.

      --
      Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
    9. Re:AI by gweihir · · Score: 1

      Pretty much the same as any longer-term hype. Always the same stupidity.

      --
      Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
    10. Re:AI by angel'o'sphere · · Score: 1

      As long as you are not a computer scientist: no you are not the one defining what AI is.
      And as most of your posts about AI are simply misleading or uninformed or simple to lay men, I'm pretty sure you are not a computer scientist, and if you probably are: you most certainly never worked in the AI sector.

      Perhaps you want to start reading here: https://en.wikipedia.org/wiki/...

      (Hint: I'm a computer scientist)

      --
      Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
    11. Re:AI by Anonymous Coward · · Score: 0

      Yes, okay? In the popular press, AI now means algorithm, or expert system, or machine learning system, or statistical model, or really whatever the writer wants it to mean. Journalists will continue to abuse the term because at the moment, it has a certain cultural cachet.

      The fact that you complain about it on *every* *single* *article* that mentions AI on Slashdot is never going to change the way journalists use the term.

  4. Fraud is a small problem? by Anonymous Coward · · Score: 0

    AI is being deployed by those who set and mark exams to reduce fraud -- which remains overall a small problem...

    A "small" problem?

    This is probably low-balling it:

    60.8% of polled college students admitted to cheating.

  5. No, it cannot by gweihir · · Score: 0

    What can raise standards is a competent, dedicated examiner, nothing else. No amount of stupid statistics and pattern-matching will help, unless the examiner is really incompetent. And stop with the "AI" nonsense. There is no "AI" in existence, the term is just a marketing-lie.

    --
    Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
  6. What's the false positive rate? by Anonymous Coward · · Score: 0

    How many folks will be accused of cheating who didn't cheat? Do they have any defense? Or are they just randomly screwed?

    How many classes do I have to take to graduate with a degree from University? How many "opportunities" will I have to be accused of cheating in each class? What are my odds for overall success? And how many hundreds of thousands of dollars am I paying for THIS ASS RAPING?

    This is going to be like the TSA bullshit. A 95% accuracy rate. 600 million passengers. 30 million false positives. Zero actual terrorists. But God help you if you're one of those 30 million false positives. "Hey, after all this hatred and abuse we finally got a terrorist. Let me see your hands! Let me see your hands! He's got something in his hands! Take him down!"

    Oy vey.

    1. Re:What's the false positive rate? by arth1 · · Score: 1

      This is going to be like the TSA bullshit. A 95% accuracy rate. 600 million passengers. 30 million false positives. Zero actual terrorists. But God help you if you're one of those 30 million false positives.

      Yes, lawmakers and voters not understanding even basic mathematics and statistics is a big problem. They look at the accuracy rate for positives and praise a high one, but not the ratio of false positives, which makes the high accuracy rate completely irrelevant. You don't hear them present it like "if someone is flagged, chances are better than 99.999% that he or she is not a terrorist".

      It's easy to create a method with 100% positive accuracy rate. Just flag everyone.

  7. Keyboard pressure? by Anonymous Coward · · Score: 0

    What browser are they using?

  8. Retarget Degree Courses by Anonymous Coward · · Score: 0

    Perhaps the better option is to retarget degree courses. Stats is 99.99999% unlikely to ever be used/remember/useful to anyone outside of a hard math degree, accounting, scientific, etc. The same is true for philosophy, the arts, economics, and all of the other bullshit that goes into a degree. People tend to cheat in classes they're not invested in.

  9. Does cheating really matter in school? by Anonymous Coward · · Score: 0

    We aren't offering gold medals here, just an education. Cheating aside, I've seen the quality of college grads, and it varies widely enough that the presence of a degree or not is largely irrelevant anyway.

    Just make the exam a measure relevant for the student, and not relevant for anyone else. Eliminate published grades entirely.

    If you want to cheat, go to town. Just don't be surprised when you get a job and don't know how to actually do anything you get fired a lot.

    1. Re:Does cheating really matter in school? by Anonymous Coward · · Score: 0

      We aren't offering gold medals here, just an education.

      They're often not offering much of an education, it's largely certification that the person is qualified, so it really is a "gold medal."

  10. Why bother after the test? by bobbied · · Score: 1

    If you don't secure the actual test taking, why does anything else matter?

    FIRST you must verify that the person taking the test is the person who's supposed to be taking the test.

    SECOND you must verify that the person taking the test has only the equipment and materials allowed by the test.

    THIRD, you must monitor the actual test taking to verify the rules are being followed and there isn't any communication between test takers going on.

    After that, mark the tests as they are and let the rest of this just be. If you want to see how effective your test day security is, sure, use such tools to get an idea if you are missing cheaters, but you won't be able to catch them this way. If you see evidence of cheating, up your security proceedures next time.

    --
    "File to fit, pound to insert, paint to match" - Aircraft Maintenance 101
    1. Re:Why bother after the test? by Anonymous Coward · · Score: 1

      My understanding is that TFA proposes a way to automate your three items. The questions are how well it works and whether it's worth the expense.

    2. Re:Why bother after the test? by bobbied · · Score: 1

      You cannot "automate" physical security activities.

      You may be able to automate checking ID's are valid, but *somebody* is going to have to look at the pictures or you are going to have to collect biometrics (again, which takes actual people to do).

      Inspecting what a person has in their possession is again not something you can automate (just ask the TSA).

      How you monitor test taking to catch rule breaking using automation? (Asking for a friend who runs a casino in Los Vegas.) It takes eyes in human heads.

      This is a physical security issue, and that's not subject to automation.

      --
      "File to fit, pound to insert, paint to match" - Aircraft Maintenance 101
    3. Re:Why bother after the test? by Anonymous Coward · · Score: 1

      I don't do any of those things. I teach a large (300 person) humanities lecture, and I give all my tests unproctored, online. Students have a window of three days in which they can take the weekly quiz, and a window of five days for the midterm and final, so that they can schedule them for a time and place convenient for them--at home eating cookies, at the coffee shop, with friends, whatever. I know that I have no way of keeping students from taking quizzes together, or letting someone else take the quiz for them.
      I do, however, have the quiz software do four things that limit students' ability to cheat: time is limited to a minute per question, the questions are randomly pulled from question banks, so no two students are likely to have the same quiz, questions are presented in a random order, and answers for any question are presented in a random order. The randomization of questions and question order is effective in keeping them from collaborating, because it takes too long for them to figure out how to collaborate on those (a control-f search would do it, but they rarely seem to figure that out). The time limit is the most important factor: a minute per question (two minutes for the students with disabilities accommodations) is plenty of time for someone who knows the answer, and, as long as the question is well-written, not enough time for someone trying to Google it.
      Students are much, much happier with this setup than filling out scantron bubbles in class. The grades have improved a bit, but not much: I still have people failing the class, despite all the tests being unproctored and online.

  11. Nope by alexborges · · Score: 1

    I don't think this is a great idea. I mean, all that money to make sure students that pay tens of thousends of dollars aren't cheating mostly themselves.... it's fucking stupid and it says more about the whole idea of modern capitalism, the illusion of meritocracy and the doubtful relevance of superior education, that about the students that would cheat themselves.

    --
    NO SIG
    1. Re:Nope by Anonymous Coward · · Score: 0

      it's a fucking crazy idea just as bad as facebook rating people. We need to rate banks and corporations (spoiler: they are the shittiest shit)

    2. Re:Nope by pauljlucas · · Score: 1

      While the students are cheating themselves, they are (for those attending a prestigious university) also diminishing the reputation of the university. If anyone could get an MIT (or pick youâ(TM)re favorite elite school) degree by cheating, the degree wouldn't be worth much.

      --
      If you reply, do so only to what I explicitly wrote. If I didn't write it, don't assume or infer it.
    3. Re:Nope by Farton · · Score: 1

      I study at the university, but sometimes I have to apply for help to essay writer. And I do not consider this to be a fraud of myself, it's just a help in studying, because problems are difficult and many, and you want to study well. It's just convenient.

  12. Find the experts by Anonymous Coward · · Score: 0

    If you want to spot cheats, look no further than the University of North Carolina athletics department. You'll have to go elsewhere for the 'raising standards' part.

  13. ncaa student athlete don't have time for class 40+ by Joe_Dragon · · Score: 1

    ncaa student athlete don't have time for class when the team needs 40+ hours a week and when travel time makes them miss class as well.

  14. they don't really expelled people that much any mo by Joe_Dragon · · Score: 1

    they don't really expelled people that much any more some times they just pass them to keep that loan $$$$ coming it.

  15. preston vue testing can do that for $50-$100 a pop by Joe_Dragon · · Score: 1

    preston vue testing can do that for $50-$100 a pop. Click hear for student loan ez-pay.

  16. Re:ncaa student athlete don't have time for class by Anonymous Coward · · Score: 0

    Why call them students then and not economic slaves?

  17. That's kind of like saying . . . by Anonymous Coward · · Score: 0

    . . . that we can use algorithms to detect cheating. No, no, it's exactly like that. 'AI' my ass. We have been using algorithms for stuff like this for a long, long time. This is not remotely new. The sooner people start replacing statements that frame algorithms in the first person and acknowledge that they are tools used by people to automate tasks, the sooner people will start taking technology like this more seriously instead of giving millennials cyber wedgies.

  18. Theirs not much point in actually sitting a test by BrookSmith · · Score: 0

    Theirs not much point in actually sitting a test if the AI already knows who is going to pass and who is going to fail.

  19. What about plagiarism by the trainer? by Anonymous Coward · · Score: 0

    ... plagiarism detectors that can scan essay answers ...

    I have to do an online 'vocation re-training' course. The essay questions come straight from the government-employee training programs. If you've done vocational training, you know it is 'write the provided paragraph under the essay question' memorization. The problem is: These training resources don't; it is general principles and rules but doesn't explain processes or how to actually do anything. This course doesn't provide anything to memorize. A google-search of the question reveals essay-answer services; which demand recurring payments.

  20. SATS anywone? by nanospook · · Score: 1

    Can AI be used to identify poorly written or "stupid" "ineffective" exams?

    --
    Have you fscked your local propeller head today?