The margin of error is usually defined as a particular way of expressing a confidence interval - see Wikipedia's definition, for example. So you are not really being asked to do anything other than re-express what the newspaper has already stated. By changing from 65% plus or minus 4 percentage points to 0.65 plus or minus 0.04 you have only changed it from percentages to proportions, by dividing by 100 (which obviously has no importance).
Looking again at your question I suspect it is actually exactly this relationship between a percentage and a proportion that is being tested here (as it is several months old I don't mind giving the answer away), not really the confidence interval issue.
Exactly what a "confidence interval" means is another matter (see @gung's comment). The usual explanation is that if you performed this exercise (sampling a large number of voters etc) many times and used the same technique each time to estimate your confidence interval, 90% of the time the confidence interval would include the true but unknown value.
BTW if only newspapers were as clear as this, stating that the margin of error is a certain number of "percentage points" - all too often they omit the crucial "points", which leaves a significant ambiguity in meaning.