Take yourself back to high school. You were out sick one day. That day you missed a massive food fight. You show up to school the next day and you want to know who or what caused the fight. Now imagine two scenarios: A) you talk to 100 people and their stories are pretty much aligned. It was Bill who started it B) you talk to 10 people and each one identifies a different person There is less certainty in situation B. This is clear. This is the same thing with confidence intervals. When we estimate the features of a population (a person's entire bloodstream, the nation's voting habits, e.g.) by using a sample (a blood test, a survey, e.g.) our confidence in that estimate is a function of sample size and standard deviation (i.e. how much the information changes/varies).

When we want to generalize a statement for a big population, it's impossible to get the data from all the members. So we take a sample of the population and use the calculated sample parameter as a proxy for the population parameter.

When we want to generalize a statement for a big population, it's impossible to get the data from all the members. So we take a sample of the population and use the use sample statistic to estimate the population parameter. But different samples have different sample statistics. Which one is a good approximate for the population parameter? So we should come up with an interval to confidently say that the population parameter lies in it. This interval is called Confidence Interval.