Fill a bowl with water and gather five objects. Before each one goes in, ask your child to predict: sink or float?
Put the prediction somewhere visible — two piles on either side of the bowl — then test it.
When a prediction is wrong, treat that as the interesting part: 'you thought it would sink, and it floated. That is surprising.' Never present a wrong guess as a mistake.
Predicting before testing is the core move of scientific thinking, and it only works if being wrong is safe. Sorting the predictions into visible piles makes the child commit, which is what turns a guess into a hypothesis they actually care about.