Comet wants the report done by tonight. "Q6 is optional, and its answers fit no category. Let's delete the whole Q6 column."
Wren pulls the Q6 answers toward him. Most are blank. A few say thank you.
Then he stops at M22. He reads it aloud: "A ramp at the side door."
"What do you notice?" he asks quietly.
Comet reads it twice. "That is not on our list of four. But it matters. Someone cannot get in easily."
Nova hovers over the page. "Would you like a hint?" she asks. "Cleaning choices decide whose words reach the council."
Comet puts the eraser down. "We keep it. We just need a new place for it in the report."
Computing innovations can reflect human biases through their data or their algorithms. Bias can enter at every level of the work.
Bias often comes from the type or source of the data collected. Collecting more data does not remove it.
Working with data is iterative. You filter and clean, look again, and sometimes go back to change a step.
If the crew had deleted Q6, nobody would know a member needs a ramp. More survey answers would not have fixed that.
| Choice | What the council would hear |
|---|---|
| Delete the Q6 column | Four categories only. The ramp request is gone. |
| Force M22's ramp into a Q3 category | A wrong total, and the ramp request is hidden. |
| Keep Q6 and report it in its own section | Four categories, plus one access need to look at. |
These choices are part of the Commons Count story.
| Statement | True or false? |
|---|---|
| Bias can enter at every level of working with data. | ? |
| Collecting more data always removes bias. | ? |
| Deleting answers that fit no category is always safe. | ? |
| Data work is iterative: you may go back and change a step. | ? |
What a week, data analyst. Tomorrow is Mission Quest: sorting with rules your whole family agrees on.