Module 0412: Student-centric generative AI
What is generative AI?
- Artificial Intelligence tool that generates content:
AI may be a misnomer
- Not artificial:
- Most AI tools are neural-net-based.
- Machine learning from observing numerous human-created content
- Humans also filter and select samples for training
- A lot of human thumbs in this pie!
- Not intelligent:
- Knowledgeable? Yes.
- Intelligent? Not yet.
- Limited critical thinking and problem-solving abilities.
- Very limited creative thinking and theorizing abilities.
Examples
Student-centric objectives
- Get employed.
- Learn skills.
- Gain knowledge.
- Qualify for advancement.
- Transfer.
Common ineffective ways of using G-AI
Computer science as a case study
Big question
- Are the SLOs applicable to G-AI or our students?
- If a student meets the SLOs but only with G-AI tools, is this student successful?
- How do we define student success?
Learning versus assessment
- Use G-AI in the learning phase, not the assessment phase.
G-AI explanation
- G-AI can help explain many concepts, but can it be relied upon?
- If there is plenty of human-generated content that is correct, then the G-AI answer is more likely to be correct.
- Does not work as well for concepts that do not have plenty of human-generated explanations.
- Can a student validate the G-AI explanation?
- Neural-net-based G-AI are probabilistic, there is always a chance of error.
G-AI Q&A
- G-AI can generate questions for faculty and students.
- Is the answer (key) correct?
- Can a student validate the G-AI answer?
A responsible way to use G-AI
- Instructor originates G-AI Q&A.
- Instructor validates and hides the answer.
- Students work on the question.
- The instructor discloses the answers and potentially evaluates the answers from each student.
Practically no way to stop students from using G-AI
- Especially for fully online (and asynchronous) classes.
- Instructors need to point out and explain self-defeating ways of using G-AI.
- There are ways to “poison” prompting text, but it is defeatable.
- Practical in some classes.
- ChatGPT accepts Markdown format as input.
- Input text from reading material, then interact as a way of studying.
- Is the result reliable?
- An example, from the beginning..
- Note how the prompt begins with “The following is some material that I want to understand. My questions will be in the following prompts.”
- The material source format is in Markdown, all the mathematical equations are captured correctly.
- In this case, ChatGPT is correct in the subsequent interaction.
- Q&A prompt: “How is a P-type transistor different from an N-type transistor?”
G-AI to generate a quick summary
G-AI as Q&A
G-AI for general study strategies
G-AI can be a patient teaching assistant
- Perhaps one of the best application of G-AI that is productive for students!
- Need the licence of the source material to allow copy and paste as part of a prompt.
- The source material can be in any format, but Markdown is the best format.
- Shameless plug: check out Tak’s OER workshop on 8/29!