|The cheeseburger/salad low carb thing at The Counter in San Diego. It isn't particularly a well designed experiment. It was pretty bad.
Often, for example, as people try things out in the kitchen, when they deviate from recipes and convention, they call that "experimenting" with the food. In fact, just bringing in anything unfamiliar, such as trying to cook the cuisine of a different culture, can be called experimentation. But in science, "experiment" is a precise and vital function in the process.
The experiment is the process in which one tests a hypothesis. In short, one should begin with a question, and the experiment should be designed to answer that question. A well designed experiment will take as many variables as possible into consideration, and will minimize ambiguity in the result. It has a definite beginning and an ending, and at the conclusion will arrive at an explanation for the question it begins with. To do this, at least two requirements are employed: controls and replication. Controls are set isolate the question being tested from other variables. For example, if you are testing that adding extra eggs to a recipe to see if it increases tenderness, then you must also perform the same recipe without the additional eggs to compare. Ideally, a control should be performed in the same time frame.
In addition, replication is needed to ensure that against the potential for error. After all, if we have learned something, it should have predictive value, and can be repeated. This is why when someone reports testing out a new diet as "feeling better", it isn't scientifically valid. This is testimonial anecdote, and is poorly repeatable - and thus, the lowest form of scientific evidence. Even though this seems to be the form people instinctively trust - and marketers exploit it with gusto.
So, the next time someone talks about being a science experiment - or experimenting with their food - ask about the hypothesis, controls and replication. Because without those, well, it really isn't an experiment.