Look at some aspect of society and find something that interests you. For instance, if you are a smoker or know someone who smokes, look into the impact of smoking on society.
Narrow your focus to something specific. What aspect of society would smoking affect? For example: How many hours in a week are lost to smoking breaks?
Much of sociology deals with statistical analysis, so ask yourself if this is something that can be tested and expressed mathematically. The inclusion of hard data in the form of numbers will give credit and authority to your interpretation and opinion. Your hypothesis, test, and results should be replicable by another researcher.
Make an educated guess about what your results will be. Your hypothesis is an assumption made from your own knowledge or common sense. Be as specific as you can, as it will make your conclusions more definitive. For example: instead of saying "I believe smoking wastes productive time through smoke breaks," make an assumption at how many hours will be lost.
Treat your hypothesis like a question, not a statement. A hypothesis is not a concrete argument; it is an assumption that you are looking to either prove or disprove. Don't be afraid of being disproved by your research.
Create a model for testing your hypothesis. Ask yourself at every step of the model creation, "Will this give me data that relate to my hypothesis?"