AI Literacy in Year 10 Biology: Enzymes
You already teach students to compare enzyme data and explain patterns in activity. The AI connection is that a classifier also uses patterns in examples, but it can produce an incorrect output when its examples or rules are limited.
Where does AI fit?
- A classifier can match patterns without understanding enzymes, so students question an output that places a statement in the wrong category.
- A conclusion is only as strong as the evidence collected: missing examples can create predictable mistakes.
- An AI recommendation can support scientific work, but it must not replace scientific judgement.
What words need pre-teaching?
- Classifier: a computer system that puts an input into a category using patterns from examples.
- Training data: examples used to help a model produce categories or predictions.
- Model: the pattern-based system produced when an AI system learns from examples.
- Human oversight: a person checking an AI output before it is used to make a decision.
Which frameworks does this hit?
- OECD AILit: Students investigated how a text classifier used enzyme examples, identified its limits and compared its outputs with biological evidence.
- AI4K12: Students examined how labelled examples and changed datasets influence an AI model's classifications and predictions.
- PISA 2029: Students compared AI classifications and recommendations with enzyme knowledge, dataset completeness and direct biological evidence.
- UNESCO: Students designed an enzyme adviser with defined inputs, predictions, warnings and a human decision point.
A short selection. A full guide maps every activity to the frameworks you choose for your region.
Your activities
Open a card to see how to run it in class.
Training Data Changes the Classifier
Plugged · 15 min
Students train and test a text classifier with enzyme statements, then compare its categories with biological evidence.
Say this to your class
"You have already learnt that temperature and pH affect enzyme activity. Now you will test whether a classifier can sort enzyme statements accurately. We will give it examples, test new statements and compare its output with the biology you already know. Treat every output as a claim to check, not as an answer to copy."
Questions to ask
- Which statement was easiest for the classifier to categorise, and what words may have helped it?
- Which test statement exposed a weakness in the training examples?
- How would you change the six training examples so that temperature and pH were both represented?
When it clicks
The concept has landed when students say that an output depends on the examples and must be checked against enzyme evidence.
Catalase Data With Missing Evidence
Start here · Unplugged · 10 min · quick win
Students alter catalase datasets and predict how incomplete evidence could change a model's enzyme recommendation.
Say this to your class
"Look at the catalase results as a scientist, not just as a list of numbers. You will change the evidence available to a hypothetical predictive model and see whether its recommendation changes. Your task is to spot when a neat pattern is weaker than it first appears."
Questions to ask
- What would a model trained only on the incomplete data fail to test?
- How does the misleading sample change the apparent trend?
- Why is a prediction not the same as a measured result?
When it clicks
The concept has landed when students explain that the same biological process can produce a different prediction when the available data changes.
The Fermentation Adviser Hearing
Discussion · 14 min
Students hold a stakeholder hearing about whether to trust an AI recommendation for enzyme temperature and pH.
Say this to your class
"You know how an enzyme investigation needs controlled conditions and evidence. We are now applying that reasoning to a workplace decision. Read your character's evidence carefully, then argue for accept, revise or reject. Every contribution must name one enzyme fact and one limitation of the AI recommendation."
Questions to ask
- Should the panel accept, revise or reject the 40°C and pH 6.5 recommendation?
- Which missing measurement most limits confidence in the recommendation?
- Who should be responsible if the recommendation is used and enzyme activity falls?
When it clicks
The concept has landed when students distinguish a useful recommendation from a decision that still requires a named human checker.
Design a Human-Checked Enzyme Adviser
Creative · 18 min
Students design paper prototypes showing how enzyme data becomes a recommendation followed by human checking.
Say this to your class
"You have examined how enzyme evidence can be incomplete and how a recommendation can still sound convincing. Now design the paper version of a safer adviser for a catalase investigation. Your design must show what data enters, what the system predicts, what warning it gives and where a scientist takes control."
Questions to ask
- Which input would your adviser need before making a temperature recommendation?
- What should happen if the data contains no result above 40°C?
- How does your prototype stop a prediction from being treated as a measured optimum?
When it clicks
The concept has landed when students place a real checking action between the AI recommendation and any change to the enzyme investigation.
Reference & support
See which activities to start with
Ranked most to least accessible. The top entries are strong entry points for any class.
- Training Data Changes the Classifier: a printed data table, with each statement read aloud before students enter or classify it.
- The Fermentation Adviser Hearing: role briefs with a highlighted claim and a sentence starter so every student can prepare a contribution.
- Design a Human-Checked Enzyme Adviser: a labelled paper prototype template, and students can explain their design orally.
Before using AI tools with students
The activities in this guide are designed to support your professional planning. Before using any AI tools with students, speak to your SLT and DPO, check tool age limits and your school’s Acceptable Use Policy, and never let students enter personal information (names, photos of themselves, school name, anything identifiable) into open or unvetted AI tools.