1. Academic Integrity
2. Assessment & Standards
3. Data Privacy & Security
4. Equity of Access
5. Skills Acquisition
6. Curricula & AI Literacy
7. Agentic Control & Governance
Did I miss anything? Mind map coming shortly.
1. Academic Integrity
2. Assessment & Standards
3. Data Privacy & Security
4. Equity of Access
5. Skills Acquisition
6. Curricula & AI Literacy
7. Agentic Control & Governance
Did I miss anything? Mind map coming shortly.
I asked ChatGPT when will SORA be available?
It replied that EU AI regulations were slowing its roll out but that VPNs allowed you bypass this. It stated that was against OpenAI policy!
Was this unethical?
Yes, it apologized and said it would never do this again!
Share!!!
6. We asked them to decide what they had learned and here is what they reported: -
a) AI is great for getting an outline of the subject.
b) AI is biased and can easily manipulate opinion.
c) You must decide what is important.
d) Exploring with AI needs time.
5. We gave them another 2 minutes. This was risky. Another round of 'hands up' but still no winner emerged. We got the volunteer to read out what we had written on the envelope. It was controversial but ultimately the students agreed it was the key fact.
4. Within five minutes, what 140 students and various LLMs had produced was an excellent if shallow account of the event. That is impressive in terms of productivity. And of course, the money was still there to be won.
Time passed, still no winner.
3. We only gave them two minutes to complete the task. The 'correct' answer, the most salient fact about this case, was written on the envelope holding the money, held by a single volunteer. Once the time had elapse, we called for volunteer answers. Hands up.
2. We used an event with which we were both familiar and which predated all the students, i.e. it happened before they were born and on another continent. There was some coverage of it recently but nothing substantial, so they had to rely on the AI outputs...
1. We ran an experiment with about 130 first year students.
We asked them to locate the five most salient, crucial or important facts about an event and encouraged them to use any form of AI they wanted. To drive participation we put €50 in an envelope...
Hi. Running a series of semi-serious experiments initially with #ChatGPT exploring the impact of AI on Higher Education. Somewhat tongue in cheek, this will be followed up by iterative experiments on the same topic. substack.com/home/post/p-...
Prompting #12
Learn.
There is nothing worse than remembering, after a series of prompts that generated less than helpful content, that you've been down the rabbit hole before!
Have you done something like this before?
Have others?
Collaborate.
Prompting #11
Returning to the ABC rule: is it Appropriate, are you using Best Practice and are you being Critical in your thinking.
You have to look at every response to every prompt using these lenses. Sometimes you don't need to use AI!
Finally...
Prompting #10
Set clear goals and stopping points to avoid excessive refinements. Stop points are really important. We are inclined to ask open questions at the start. All good. You must start to close them, to funnel down to the topic.
Prompting #9
Test prompt variations systematically and track effects for better results.
You can use Excel to track this, putting large outputs into documents and linking to them directly, with a summary of how the results differ from past iterations.
Prompting #6
ABC: Use neutral language, test for bias, and explicitly constrain tone or content. If you're prompts are quite conversational or loose, the results will reflect this.
If you ask for "enormous", you will get whales!!!
Prompting #5
Balance examples with general instructions; diversify examples for flexibility. "War stories" are powerful tools in communication. Ask it to locate examples of [topic]. Ask it to look on specific social media channels if desirable.
Prompting #4
Direct the model to "Be concise" or "Avoid repetition" and adjust temperature and language. AI is criticized for writing in "LinkInEse" - Direct it to not use this kind of language. Insist on accuracy and full references /links.
Prompting #3
Restate context or goals in each iteration to maintain focus and consistency. State who the audience is and their level of expertise. Specify the language: can you use acronyms or do you need to explain every term?