In the staff room, a colleague tells Ben she lets a chatbot write her weekly family newsletter. "I just send it," she says. "It always sounds great."
Ben thinks about that all afternoon. It may sound great. But who checked the dates and the names?
Maya has a similar worry about meal plans, and Rosa about her shop posts. How much should the tool do, and how much should they?
The US Department of Education report from week 1 offers a simple picture for AI.
On an electric bike, the rider is fully aware and fully in control. The motor just makes the work lighter.
A robot vacuum is different. It works on its own while nobody watches.
The report says it hopes AI will be more like the electric bike, and less like the robot vacuum.
| Question | Electric bike | Robot vacuum |
|---|---|---|
| Who steers? | You do | The machine does |
| Who watches the road? | You do, the whole time | Nobody, most of the time |
| Who decides when to stop? | You do | The machine, or a timer |
| What does the machine add? | Less effort | The whole job |
The same report makes several recommendations. The first one is "Emphasize Humans in the Loop."
It says teachers, learners and others need to keep the power to decide what patterns mean and what to do.
For you, that means the tool can draft and suggest. You decide what is true, what fits and what gets sent.
If keeping control is so sensible, why do people stop checking? NIST, the US standards agency from week 1, names one reason.
People can over-rely on generative AI. They may see its output as better than content from other sources, even when it is not.
NIST calls this automation bias. It means trusting a machine too much, and NIST says it can make other risks worse.
In week 2 you saw that these tools can state false things with confidence. Automation bias is what lets those errors slip through.
| Statement | True or false? |
|---|---|
| On an electric bike, the rider stays fully in control. | ? |
| The report hopes AI will work more like a robot vacuum. | ? |
| Automation bias means trusting a machine too much. | ? |
| NIST says automation bias can make other risks worse. | ? |
| A polished, confident answer has already been checked. | ? |
This week you practice two habits: ask well, and check always.
They are methods this course suggests. No source here proves a set recipe that always works.
They do one thing reliably. They keep you on the bike, with your hands on the handlebars.
Sources: US Department of Education, AI and the Future of Teaching and Learning (2023) · NIST AI 600-1, Generative AI Profile (2024)
Tomorrow: how to write a clear request in four parts.