Sometimes, the 'hard' sciences are paid so much attention and done ad-naussiem with little variation from year to year (how many baking soda volcanoes and bread mold experiments are there?)
There are other, 'softer' sciences - such as the social sciences. You could, for example, do an internet experiment probing how different communities react to different topics or polls. ... You could make it simple: "More males role-play female characters in video games that allow gender selection than females role-play males." - then test the hypothesis through a series of polls.
You could even form distinctions between different communities. "Anime forums are likely to see a higher percentage of males role-playing females than tech forums (such as Anandtech, AMD developer boards, etc).
Now - with such polling, you would have to acknowledge the limitations of polling accuracy and discuss limitations or artifacts in your sampling methods (as well as what measures you took to reduce such things) - but you can end up with a very interesting science project that will not be a rehash of the same-old.
You could also try more personal things - such as greeting people in a store or a range of environments, documenting the reactions of people as you varied your clothing or manner of speech.
It's important, with each of these, that you research doing proper controls and how to do a proper discussion and conclusion of your project... but these are often low-cost projects that are also rather straight-forward and simple (they just require the initiative and a level head to set them up).
Or you could do what I did and build a gauss rifle and/or rail gun.
