Although valuable insights can be gleaned from observing people, the findings are unreliable at best due to humans' unpredictable nature. A psychologist can make general assumptions about how a person might react in a given situation, but there is no guarantee that every person will react in the same way.
Observing people through controlled experiments is also unreliable, because it would be difficult to replicate the ideal setting and situation of an experiment out in the real world, and vice versa. A person might behave differently during an experiment in which they are being observed as opposed to facing the same situation in real life. These findings are not very useful in predicting human behaviors.
Psychology can't be considered a science if its entire base of knowledge is muddled with such inconsistencies.