Itd AI guidance - #1950
Conversation
✅ Deploy Preview for cyf-curriculum ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
✅ Deploy Preview for cyf-common ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
|
Related ticket: #1901 |
illicitonion
left a comment
There was a problem hiding this comment.
This all LGTM, left a few suggestions, the side-by-side CV may be a little work to put together but I think is probably really worthwhile? I've definitely had a lot of experiences of pointing out to a trainee (particularly on ITD when we used to have them generate CVs/cover letters with GenAI) how bad their AI-generated cover letter was
| Success at CYF is not about using AI to rush through the course and tick all the boxes. It's about building a deep understanding of the concepts we teach for yourself. Using AI to do the work for you wastes everyone’s time. You will end the course without the coding skills you wanted (which means you won't get a job) and our volunteers will have wasted their time marking AI work. | ||
| This means that all output from AI is potentially untrustworthy, we should keep a critical eye on the output of AI tools. | ||
|
|
||
| ### Why do we need to be careful as ITD trainees? |
There was a problem hiding this comment.
I'd be tempted to split this into two sub-sections:
- We need to build understanding
- Sometimes it lies
I think the "we need to question and verify and explain" could do with being split into those two sections too:
- In the "building understanding" section we can talk more about how we should be able to spot problems in AI output, and we should be able to delete AI output and write it ourselves because we know how.
- In the "sometimes it lies" section, we can talk a bit about looking up other sources
|
|
||
| You need to exercise CAUTION and avoid the following situations: | ||
|
|
||
| - Letting AI write your CV |
There was a problem hiding this comment.
| - Letting AI write your CV | |
| - Letting AI write something for you (e.g. your CV) |
And let's explain why this is bad - that it will generate generic CVs which don't sell you, which aren't personal, that use filler words, that make things up about your background, etc.
Maybe we could even include side-by-side a bad AI generated CV and a good personally written CV, and give trainees a prompt for some criteria to use to compare them, and have them submit their comparison? (On average our ITD trainees don't know what a good CV looks like, so I think we would need to give them criteria to evaluate from)
There was a problem hiding this comment.
Thought dump after two weeks of chewing on how to make AI CV creation make sense
Rn the learning objectives are going in the direction of:
- Highlight the ways AI can make mistakes or go wrong.
- Highlight the disadvantages of AI-generated work.
- Identify good and bad uses of AI for learning during the ITD.
- Create a Google Doc with answers to questions about AI usage.
Point 2 is where CV comparison could come into play. Issues I'm running into atm:
Upskilling ITD folks to know what a good CV looks like and how to spot issues. A lot of content for a page focused on AI.
Workaround: I am delegating upskilling trainees in what a good CV looks like by pointing to the guide and evaluator tool. It feels like a lot to explain in the prep.
I can currently force AI to hallucinate if I give it a more mid/senior role and a junior "profile" and ask it to generate a CV for it (just getting it to write a CV based on a profile was actually hard to make look bad, common LLMs seem to reasonably make a CV on its own - ChatGPT is fluffier than claude for sure).
Now the issue is, getting people to spot the hallucinations and fluffed up language is hard when they have no idea about the language, ideas, engineering concepts mentioned.
Workaround: I can try to make the profile someone who has warehouse/taxi/shop assistant experience and go for a more customer service/intern/office role. But again hard to tell if people will have enough understanding of the skills to spot the hallucination.
If the core q is "why do I bother writing a CV myself" we can drill in these advantages:
- accuracy in a CV, hallucinations avoided
- English lang skills
- Understanding the contents of the CV for interview fluency
And an alternative good use of AI might be finding out more about what a good structure of a CV would be for a uk-based SE, and a bad use would be "write a CV for me" (and why these are good and bad).
I'm still chewing on this but needed to put a bookmark in where I'm at with this
…nto itd-ai-edit
Co-authored-by: Daniel Wagner-Hall <daniel@codeyourfuture.io>
…nto itd-ai-edit
to explain how AI can make mistakes and how we should prioritise understanding
to show drawbacks of AI to generate written work like CV
to correct CV review guidance and various wording across page
…nto itd-ai-edit
LonMcGregor
left a comment
There was a problem hiding this comment.
Overall I like these changes. The "how this works" changes look good to go. I've got some comments on the steps and AI stuff
|
|
||
| When people [sign up to study](https://codeyourfuture.io/itc/) on our website, they are invited to begin our Intro To Digital (ITD) Course. This course lasts one month, and runs 5 times a year. | ||
|
|
||
| Learners complete the course on our [Guided Learning platform](https://course1.codeyourfuture.io/) by completing a sequence of steps. Each step is explained clearly in B1 English. The text is reproduced on this site in the [Steps](./steps) section. |
There was a problem hiding this comment.
Testing this on the deployed site, the steps link goes nowhere right now. Should that take you back to the main page where the steps are listed?
|
|
||
| The first makes you stronger. The second leaves a gap that will show itself later, often at the worst moment, like a job interview. | ||
|
|
||
| **Ask yourself: “Will I need to ask AI the same question tomorrow?”** |
There was a problem hiding this comment.
Is this meant to be instruction 3 following 1 and 2 above?
|
|
||
| Look for **verified sources**: | ||
|
|
||
| - Ask AI to cite it's sources, where has it found it's information? |
There was a problem hiding this comment.
| - Ask AI to cite it's sources, where has it found it's information? | |
| - Ask AI to cite its sources and then check you trust where it found its information. |
| And what happens when AI writes a CV for us: | ||
|
|
||
| - You don't build English language instincts, you only read an example of written English. You don't make your own mistakes so you don't learn what is right or wrong. If you get a perfect instance of English generated in the first go, you are less likely to know what mistakes to correct in the future when you write your own words. | ||
| - You don't build a mental map of your CV through structuring it yourself. Harder to recall your own CV in interviews. |
There was a problem hiding this comment.
| - You don't build a mental map of your CV through structuring it yourself. Harder to recall your own CV in interviews. | |
| - You don't build a mental map of your CV through structuring it yourself. This makes it harder to recall your own CV in interviews. |
|
|
||
| #### How to check if you're using AI right | ||
|
|
||
| Ask yourself: **Can you do this task next time without using AI?** |
There was a problem hiding this comment.
I think this is repeating what was already suggested earlier, but at this stage this is probably a good idea.
| Which of the below are NOT appropriate actions by the trainee? | ||
|
|
||
| 1. Asking volunteers for guidance at a session. | ||
| 2. Giving an AI tool their assigned problems and asking it to solve them for them. They then copy and paste the output, making no edits. |
There was a problem hiding this comment.
| 2. Giving an AI tool their assigned problems and asking it to solve them for them. They then copy and paste the output, making no edits. | |
| 2. Giving an AI tool their assigned problems and asking it to solve them. They then copy and paste the output, making no edits. |
There was a problem hiding this comment.
This is the longest of the steps they will need to read. Would it be possible to split up into separate pages? There are lots of sections, some of these should be grouped together, but I'm not sure of the best way
- What is AI
- How to use it
- Why we need to be careful
- The acceptable use policy
- The CV exercise
- the questions
Maybe 1,2 & 3,4 & 6,7
Breaking these up a little would be a different layout to the other steps, but might help to get the point across without being a huge wall of text.
There was a problem hiding this comment.
Reframing these to be "search or ask AI" is a good change.
|
|
||
| #### 2. Make a Plan | ||
|
|
||
| Start by thinking about what you need to know to solve this problem. What might be a good first question to ask the AI? How can you avoid having the AI do your thinking for you? |
There was a problem hiding this comment.
Is it worth having an extra question like "Do you already have an idea of where to start before you ask AI?" To try and get them in the mindset of not going to AI first all the time
There was a problem hiding this comment.
For this step, it is definitely pushing people to using AI, but at least not for using code. I would have added some sort of "escape route" for anyone who felt confident trying the task themselves without using any AI at all, but that might be out of scope for ITD and would need to be balanced against people who might try to cheat and pretend they did it themselves

This PR will
Areas review is especially needed