The first clips will create a dopamine loops based on created emotions, it can be whatever, even a disgust. Then the loops will get connected to the physical movement, the swipe. It can be a sound (pornhub, netflix, phone vibration).
Then slowly the loop will move before the swipe, so the expectation of the swipe and the potential emotion will be enough to release the dopamine.
But over time the amount of dopamine decrease and you’ll have to swipe more and more often to squeeze something. 
And the algo just makes sure to keep you in doing that based on their retention metrics.