The scenario that would carry the least amount of bias is:
randomly selecting 5 students from each of her class sections to use as the sample.
This method helps to ensure that the sample is representative of the entire population of students in her science classes, reducing the risk of bias that might come from only selecting top-achieving students, only male students, or only students from specific periods.