AIYINGFrom AI Awareness to Accomplishment
Showing posts with label AIYING Practice Cards. Show all posts
Showing posts with label AIYING Practice Cards. Show all posts

Friday, November 20, 2026

The Explanation That Felt Like Understanding — How to Test What You Learn with AI | AIYING Practice Card 003



Ravindra Dastikop with ChatGPT

We were taught to study worked examples until each step made sense. Now that AI can help with that, what should we learn to do?

An AI tutor explains a percentage problem. Every step makes sense as you read it. Then the numbers and the situation change, and you have to decide where to start. That is the moment Practice Card 003 asks you to rehearse.

Unlearn this. Moving on as soon as an explanation feels clear. Start learning this. Apply the same idea to a fresh problem and explain why your method works.

Why the card asks for an independent attempt

In a high-school mathematics field experiment, students using unrestricted AI improved during practice but performed worse on an unaided assessment than students who had practised without AI. A tutor designed to give hints largely mitigated that downside. This gives us a reason to test unaided performance in Card 003. The study tested tutoring, and the card’s effectiveness still needs a pilot. Bastani et al., PNAS, 2025.

In separate experiments, Andrew Butler found that repeated testing improved later retention and application to new questions compared with repeated study. This helps explain why the card changes the problem after feedback. A single fresh problem gives you an immediate check on whether you can reuse the idea. Butler, Journal of Experimental Psychology: Learning, Memory, and Cognition, 2010.

Try it in ten minutes

  1. A price drops from ₹100 to ₹80. What percentage increase restores the original price? Write an answer and name the amount you used as the base.
  2. Share the problem and your attempt with AI. Ask it to check your reasoning and ask one question about the base you chose. Revise before asking for the solution.
  3. Without AI or hints, try a new case. A club’s membership fell from 60 to 45. By what percentage must it grow to return to 60? Apply your percentage to 45 to check whether it restores the original membership.

Teachers can use this as a short classroom exercise for students who already know basic percentages. Have learners compare their choice of base after everyone has attempted the club problem. The card also works for independent study and belongs in the Check Before You Trust AI workshop pack.

How will I know I am learning this?

I can solve a fresh problem without hints and explain which amount I used as the base and why.

Thursday, November 19, 2026

The Script That Passed Once — How to Test AI-Generated Code | AIYING Practice Card 002

AIYING ANALYSIS · PRACTICE CARD 002


The Script That Passed Once

Ravindra Dastikop with ChatGPT

We were taught to get our code running first. Now that AI can help with that, what should we learn to do?

Imagine an AI-generated script that combines weekly sales spreadsheets. It finishes without errors and produces a plausible total. This week, however, the same sale appears in two files. If each sale ID should count once, can you predict the correct total before you run the script?

Unlearn this. Treating a successful run as proof of correctness. Start learning this. Turn a business rule into a test by choosing inputs and working out the expected result independently.

What the research supports

Broader tests catch errors that initial checks miss. Jiawei Liu and colleagues expanded the HumanEval coding benchmark’s test cases roughly 80-fold. In their evaluation of 26 models, the expanded tests caught incorrect generated solutions that had passed the original tests. That finding concerns a benchmark of programming tasks. Card 002 applies the same testing idea at spreadsheet scale by choosing an input that distinguishes counting rows from counting sales. Liu et al., NeurIPS, 2023.

Start with the requirement. NIST’s developer verification guidance recommends black-box tests derived from specifications or requirements. In this exercise, the requirement is that each sale ID counts once. You can check the result without inspecting the implementation. If two rows share an ID but show different amounts, agree on what the script should do before writing that test. Flagging the conflict for review is one possible rule. Black, Guttman and Okun, NIST IR 8397, 2021.

Check what learners can do on their own. In a high-school mathematics field experiment, students with unrestricted AI access did better during practice but performed worse on a later unaided assessment than students who had practised without AI. A tutoring version designed to provide hints largely mitigated that downside. The finding gives Card 003 a reason to include an unaided check and informs the fresh shipment case in Card 002. A workshop pilot would help establish whether Card 002 improves independent test design. Bastani et al., PNAS, 2025.

Try it in ten minutes

  1. On paper, calculate the expected sales total. File A contains A101 for ₹1,000 and A102 for ₹500. File B repeats A101 for ₹1,000. Count each sale ID once.
  2. Ask AI to review your reasoning and pose one question before giving the answer. Revise your test by recording its inputs, rule and expected total.
  3. Without AI, try the shipment case. File X lists S1 for 8 units. File Y repeats S1 and adds S2 for 5 units. Each shipment ID counts once. The script reports 21 units. Calculate the expected total and explain the discrepancy.

No coding setup is needed. In a workshop, ask learners to compare which rule they used after they finish the independent attempt. Share the card with someone learning to review generated code or spreadsheet automations.

How will I know I am learning this?

I can predict the correct result for a new test and explain why a successful run can still be wrong.

The Client Email — How to Spot Implied Promises in AI Drafts | AIYING Practice Card 001

AIYING · Strategic AI · Unlearn This. Start Learning This. · Practice Card 001



Ravindra Dastikop with ChatGPT

We were taught to reply quickly to show we were responsive. Now that AI can help with that, what should we learn to do?

The Client Email

A client wants the final report on Friday. Your team can send a reviewed summary on Friday, and the full report is scheduled for Tuesday. A draft says, “We’ll aim to get it to you by Friday.” If the client reads “it” as “the final report,” what will they put on their calendar?

Practice Card 001 gives you a ten-minute rehearsal for that decision. It suits enterprise teams and project leaders, and anyone who sends updates on behalf of a team can try it. Bring paper, a pen, and an AI chat tool for feedback. The scenario is fictional and gives you all the facts you need.

Why this card exists

Unlearn this. Drafting the reply before checking what you can promise.

Start learning this. Spotting implied promises and making commitments you can keep. Read the email from the recipient’s side. Identify the deliverable and date they could reasonably plan around, then match both to what the team has confirmed.

What the research supports

  • AI can help with the writing. In a randomized experiment involving 453 professionals, participants with ChatGPT took 40% less time on average and their output was rated 18% higher on the writing tasks studied. Those results support using AI to help with drafting. The experiment did not measure whether teams fulfilled commitments made in the writing. Noy and Zhang, Science, 2023.
  • Verification deserves its own practice. In a survey of 319 knowledge workers, participants described checking information and integrating AI output as part of how they thought critically while using AI. These are self-reports. Card 001 turns one such review task into something observable: can the learner match every promise to a fact? Lee and colleagues, CHI, 2025.
  • Why the series includes an independent attempt. In a high-school mathematics field experiment, students using unrestricted AI improved during practice but performed worse once AI was removed. A tutor designed to give hints mitigated that downside. This gives us a reason to test unaided performance in Card 003. The study tested tutoring, so the card’s effectiveness still needs a pilot. For Card 001, the unaided follow-up is a design choice inspired by the same question: what can the learner do on their own? Bastani and colleagues, PNAS, 2025.

Try it, then share it

  1. Draft a two-sentence reply to the Friday request on your own.
  2. Ask AI what the client would expect to receive and when. Revise any wording that suggests more than the facts support.
  3. Try the card’s fresh case without AI. A partner wants a full dataset on Monday, when only a sample is available. The full dataset is due Thursday. Underline each promise in your reply and check it against the brief.

For a team session, have two people read the same draft independently and write down what they expect to receive and when. If their answers differ, compare the words that led them there. Pass the card to a colleague who sends delivery updates or practises them with new team members. In the Check Before You Trust AI workshop pack, Card 001 is the exercise for checking commitments.

How will I know I am learning this?

I can spot an implied promise and write a reply whose commitments match what my team can deliver.

Accomplish with AI

Powered by Blogger.

Featured post

The Explanation That Felt Like Understanding — How to Test What You Learn with AI | AIYING Practice Card 003

Ravindra Dastikop with ChatGPT We were taught to study worked examples until each step made sense. Now that AI can help with that...

Search This Blog

List of blogs

Search This Blog

About

Translate

Instant Stories

Read in Your Language

Test

share this

ShareThis

Popular Posts