Assumptions, or cognitive inferences, are what separates humans from robots. One very salient instance of this is our ability to see a man and his shadow against a wall, and not perceive that there is actually another physical object next to the man. Robots need to be programmed an infinite number of rules to overcome just this problem which our brain easily solves by utilizing assumptions that have been formed based on our experiences and through learning
One very interesting way of teasing out these assumptions is by means of optical illusions. Optical illusions fool us because they violate our assumptions about what we see. A really good one I'd recommend is this illusion by Edward H. Adelson.
What is special about this, you might ask? Well, Tile A and Tile B are objectively the same colour.
Look again. It might be hard to believe at first, but it really is!
And to prove it (I couldn't believe it myself initially), I did the following. I created a brownish-green oval, copied it so that there are exactly two same coloured ovals, and shifted them into the tiles.
Amazingly, the two ovals appear different accordingly.
To shortcut the process above, here's probably what's going on.
The bar in the middle is really a uniformly grey bar.
What's happening is that our mind cannot divorce the effect of shadows from our perception. As long as the picture shows the green cylinder casting a shadow, the 'shadow assumption' module of our brains gets activated and the things in relation to it will be affected. A robot should typically see Tiles A and B to be the same.
Our assumptions fill in the gaps so that our perception of the world becomes seamless and efficient (and it doesn't feel like we're constantly bombarded with stimuli).