Say we got data from the experiment or computational results.
However, instead of using the line to connect data one by one. We use the spline to interpolate the data to make it look more smooth.
Also, we can use the different curve to interpolate. What curve should we believe? 
It's not quite clear what you mean by "cheat", but it certainly is misleading. You're giving a false impression of how much certainty you have. If you had splines for which the number of points in each spline were significantly greater than the order of the spline, that would be a different matter, but otherwise splines are pretty much the definition of overfitting. The curve is doing something different in the interval [0,2], and three points isn't nearly enough to say what that something different is, so drawing a smooth curve as if you can say anything with any degree of certainty as to what's going on in the intervals (0,1) and (1,2) is highly misleading. The data point at x=1 seems to be a clear outlier, and you should look into what could be causing it. As it is, simply presenting a scatterplot would be the most honest presentation.