[La-cgs] Update on AI Usage case

[B. Ramu Ramachandran]

Sanjay:

Thank you for this info. I agree that "all of the above stems from considering AI usage as an impediment to learning." We need to look at generative AI based on large language models as the "calculator for writing." My view on AI usage may be stilted because of my discipline -- most scientists suck at writing and so any tool that helps one express their ideas clearly (spell-check, grammar check, AI) is always welcome. The trick is to know enough about the subject to recognize where AI goes off the rails -- like a pilot who knows when to take over from the autopilot.

Speaking of going off the rails, I am attaching an example that a colleague shared. The blue highlights denote AI-generated text (More than 80% of this report was AI-generated). Note the first sentence of the "Theory" section where Ohm's Law is applied to a vibrating string. Wow! This student did not learn enough physics to recognize that Ohm's law relates voltage, current, and resistance in a conductor, and has nothing to do with tension, mass, and wave propagation in a string!

Another problem -- the symbols for density and speed did not copy over from the AI window to MS Word. The student was clearly too unengaged to notice that! Notice Eq. 1-1 where there are no symbols -- not even the tenstion T!!!

I believe Jason had asked for examples of where AI goes wrong -- here is one! In the hands of a careless user, even the best AI is only of limited value 😂

Best regards,
Ramu

On 5/1/2024 10:46:15 AM, Menon, Sanjay <sanjay.menon at lsus.edu> wrote:
Hi all,
 
Some of you might recall, that during the AI discussion I had mentioned submitted student work with Turnitin AI scores of over 50%.  I spoke to our Dean of Students who confirmed that Conduct Board complaints based on AI usage is on the rise. Here are some other points she mentioned that may be of interest to this group:
 
* It appears that a high Turnitin AI usage score by itself may not be sufficient to prove AI usage.  Unlike traditional Turntin reports that provide sources that can be verified, the same is not possible for AI outputs that are generative. 
* The board is now asking faculty/students to provide writing samples for comparison.  This may not be of much help in 100% online classes where the writing sample could also be AI generated.
* There is an increase in “confessions”.  More students are admitting AI usage when confronted with the AI usage score.  This raises an equity issue as those who deny usage are more likely to get away with it if AI use was prohibited by the professor in the syllabus.
* Professors’ policies on AI usage widely varies across campus with varying degrees of AI usage allowed.  Students can claim confusion when taking multiple classes with varying AI policies.
 
As I mentioned during our session, all of the above stems from considering AI usage as an impediment to learning.  If we were to adopt the approach of “AI as an aid to learning” or “AI usage as a required skill”, then we would need to rethink and reorient our teaching and evaluation methods, including for our thesis/dissertation students. You might recall our discussion on the value of doing a literature review using AI, or having AI generate code in a programming class.
 
As this issue is unlikely to be resolved anytime soon, I suggest we have it as an agenda item for future meeting.    
 
Sanjay
 
Sanjay T. Menon, Ph. D.
Dean of Graduate Studies
Professor of Management
Director of India Studies
Louisiana State University Shreveport
Office: (318) 797-5247     Fax:     (318) 798-4120
www.linkedin.com/in/sanjaytmenon [http://www.linkedin.com/in/sanjaytmenon]
 
One University Place
Shreveport, Louisiana 71115
[LSUS_Secondary_RGB Purple-Gold_email]
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.latech.edu/pipermail/la-cgs/attachments/20240501/a4bec533/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.jpg
Type: image/jpeg
Size: 1090 bytes
Desc: not available
URL: <http://lists.latech.edu/pipermail/la-cgs/attachments/20240501/a4bec533/attachment-0001.jpg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: AI-hallucination-example.png
Type: image/png
Size: 181611 bytes
Desc: not available
URL: <http://lists.latech.edu/pipermail/la-cgs/attachments/20240501/a4bec533/attachment-0001.png>