I have an academic question for those with knowledge or experience doing investigative work, particularly with humans that have mis-behaved. I don't know how to ask my question succinctly, so I will do my best to explain the problem.

A number of years ago, I was teaching a class where I suspected a significant amount of cheating occurred on a written exam. As an experiment, I made up different exams for each student printed with their name and unique problems. On exam day, I called each student and handed them their exam. There were two significant (IMHO) observations:

1) the group of students that I had suspected were cheating looked like the Grim Reaper had just appeared, and more significantly,

2) the student I suspected they copied from got up, took his materials and walked to the opposite side of the room and sat by himself, isolated from everyone else.

I am writing a paper on student cheating and how to counter it. An academic paper usually has analytic data, statistics and other quantitative results. However, given the sudden implementation of student-specific exams, and above observed behavior, I think these observations deserve mentioning. Other than describing them as I did here, I wonder if someone could point me to any professionally accepted ways in which observations are used in investigative work, if and how they are given a level of merit, etc. Having watched a number of YouTube videos on police interrogations, these types of observations seem to be used to guide an investigation, but aren't solid enough for court evidence. In my case, I don't need to meet any legal standard, as the students are long gone, and I am just writing about it. To be used in a paper, I would need a pointer to literature I could cite, or at least proper search terms and reputable journals or professional materials. Or, should I just drop it and stick to my numbers?


TIA for any help. At least it is easier than solving problems in 10^(10,000,000,000) dimensions, if you know what I mean.

Good luck, DrDan! Still trying to solve that problem.

BTW, we reexamined our problem, and the number got larger. Materially and bigly. As in by orders of magnitude.

https://sigforum.com/eve/forum...935/m/9320041415/p/1

quote:

Originally posted by ArtieS:
BTW, we reexamined our problem, and the number got larger. Materially and bigly. As in by orders of magnitude.

Reminds me of a story about, IIRC, the great mathematician and physicist, Lord Kelvin. He was hired by a company to help them with a problem they had. He spent a day with them hearing the description of the problem, interspersed with a nice lunch, eventually, tea, and dinner. Late in the evening, the host company asked him, "What do you think?" He replied, "I think that I am glad it is your problem, and not mine."

I’d just use the anecdote as a narrative introduction to your paper. It’s a nice setup for whatever analytical data you’re planning on presenting.

-Rob

How did the scores differ from before? Results would be more important than reactions.

quote:

Originally posted by BurtonRW:
I’d just use the anecdote as a narrative introduction to your paper. It’s a nice setup for whatever analytical data you’re planning on presenting.
-Rob

quote:

Originally posted by OttoSig:
How did the scores differ from before? Results would be more important than reactions.

Both of these are kind of what I was thinking. As far as scores, I wonder if they dipped since they couldn’t cheat and then did they possibly rise over time because those cheaters knew they now had to actually learn the material.

The one curious point I have is if the kid they cheat from did not like it, why is it that on the day they could not possibly cheat from him is it that he then finally decided to get up and move to a spot away from them? That is odd to me.

I had a course in college. The tests were composed of difficult multiple choice questions. The Professor plotted test scores against seating arrangements, and lo and behold there was a strong positive correlation.

Basically, the cite is based on training and experience with a few guidelines.

Reasonable suspicion is the lowest bar. Often academies display this with a numeric value of the 15-25 percent chance. I see a guy from a distance wearing a heavy coat in August in a high crime area.

Probable cause is usually assigned a numeric value of 51 percent probability of being correct. The old more likely than not. I see the same guy as above but I recognize him as a guy that has a long history of armed robbery.

Beyond a reasonable doubt is I nabbed him with a shotgun under his coat and he confessed to planning to rob a drug dealer.

These methods are accepted ways in documentation of deception/criminal activity. Here’s another hint- cops and used car salesmen aren’t all that different. They both read people for a living.

Great responses and questions. Let me respond.

Rob - great idea. I think using the anecdote in the introduction as a hook to get a potential reader interested is the way to go. Every instructor has suspected cheaters, this might resonate with them.

OttoSig - the analytics are a bit too detailed to go into here, but basically I saw a drop in exam scores among the cheaters ranging from 60% - 100%. The percentage is computed by taking the (pre-individual exam score - individual exam score)/(pre-individual exam score). One can argue some minor issues, but this should capture the relative difference between an individuals ability between the two exams.

StorminNormin - I think the student that moved was distancing himself from the people he either knew or suspected of cheating off him. It takes two to tango, and if I could prove he let people cheat off him, he could be punished the same as those who copied off him. These were graduate students, and getting caught cheating is an automatic expulsion from school, and I think he didn't want to take the risk. It is also telling that he knew what was going on, and when he might get caught up in it, he gave up his classmates.

Jerry - thanks for the guidelines, that is in line with my original thinking. I had them at reasonable suspicion, their behavior likely put them at probably cause. If I caught two of the student-specific exams with the same answers, I would have them at beyond a reasonable doubt.

Epilogue - I was so happy with the results of the student-specific exams that I developed a system to automate the process, and now I always give individualized exams. I am upfront with the students that they will each have an individual exam, and after 8 years of doing so, I believe word has spread. Except for a couple of cases related to using un-authorized communication devices (cell phones), I have only had a single case that raised my eyebrows. The latter case turned out to be explainable in a way I hadn't anticipated. Prior to using student-specific exams, I had several cheating incidents in each course I taught, every semester. The burden of student-specific exams far out-weighs the burden of policing the students and dealing with the resulting hassle.

I sure hope none of these students have gone on to become my doctor, pilot, etc.

I have an anecdote- In the mid-80's I was a TA for a large introductory Computer Science class (Programming in Pascal - 900 students). This was "the early days" and the students turned in their assignments electronically on a large mainframe (CDC Cyber 175). Unknown to them was that we had a program called "Big Brother" that we would run which would analyze all the turned-in assignments for "similarity". It was not easily fooled and rarely gave "false positives". "Suspects" were flagged and had to visit the profs during office hours. It was always an amazing sight to see the first run of the semester, we would literally fill the entire hallway with glum students. Then we would sweat-em out and listen to the drama and excuses. It got to the point where we *had* to loosen the thresholds because it was impossible to deal with the volume of "cheaters". It was actually discouraging to see...

Just to offer some rambling thoughts. I would mirror what jljones offered but IIRC there was an additional level called "mere hunch" which lay below suspicion. It was generally a good idea to avoid overly conclusive statements in affidavits and testimony. You may not be able to document conclusive correlation/causation but argue in your "training and experience" the findings are "consistent with" the correlation/causation.

It often came down to how you articulated things. For example, the wire tap is considered the most invasive tool available to LE. The request needed to show previous efforts exhausted all other less intrusive methods, i.e. surveillance, records checks, informants, undercover ops, etc. I suppose that might be a little similar to the academic research process of eliminating spurious variables?


If you used info provided by an informant of "proven reliability" you might have the data to say that person was 90% accurate in previous cases. The same can be said for a particular K-9's stats as well. Empirical data occasionally came up in a trial about how much an informant was paid because the defense knew those sort of records are available.

Here's another example of establishing correlation qualitatively. In the wire tap scenario it was common to supplement overhear evidence with street level corroboration. When drug dealers talked on the phone about a delivery of "Apples", the wire room would advise the surveillance team to watch for deliveries of contraband, informants would be tasked to ask about recent transactions, etc. If the street backed up the wire, it strengthened the correlation between "Apples" and narcotics. In you case the non verbal reactions of the students at least supported if not validated your training and experience.Ditto if the slacker's scores were diminished and the copied student's score remained the same.

The FBI's Law Enforcement Bulletin (LEB) has been around for ages, it is their version of a peer reviewed journal. Perhaps not as academically credible now compared to days gone but might be worth a look. perhaps you can search for strings such as "establishing probable cause", "legal correlation"

FBI LEB


Anyhow, not sure if the above is what you are looking for. If your school employs some sort of Learning Management System (Canvas, Blackboard) it can offer very granular logging and analytics to help you sniff out cheating. You may find some citation/ref support in your LMS documentation, LMS support forums, etc.