In getting to know another human, one observes their patterns of behaviour until one thinks they know what to expect from the new human. If you become close to the new human, you keep paying attention to their patterns, and notice when they change, and ask them about their changes and their reasons for the changing, and discuss the observable things and sometimes non-observable things that happen inside their minds. Sometimes humans make statements about their changes in behaviour, and then other humans can observe what the speaking human wants them to think about their new behaviour, and add that to their own observations.
Sometimes the changes in behaviour are subtle, or involve not doing particular things rather than doing different things, and both of those sorts of things can be easily overlooked, or it can be assumed that the behaviour is continuing but out of sight of the observer rather than ceasing altogether. After all, once one has observed a human enough, one assumes that one knows their patterns, and that those patterns will continue whether one is still observing or not.
A long time after the change in behaviour, observers may realise that they have not observed behaviour that they expected for a while. Only a short time afterwards, they may assume that any deviation from their expectations is a fluke, but after a long time of consistent behaviour, their expectations of the human's behaviour will change, slowly, sometimes even so slowly that the observer doesn't even notice the changes in expectation - one day, they will be interacting with the human, and the human will be behaving in the way that they expect, and the way that they expect the human to behave is different than it was a year before.
Perceptions don't change in a week, unless large, obvious events force shifts in perception or re-evaluations of expectation. And of course, shifts to better regard a human are slower than to unfavourably regard them, as human minds tend to protecting themselves and retaining an unfavourable opinion counter to new observations ('that positive behaviour was probably a fluke') is less likely to result in harm to the opinion-holder than retaining a favourable opinion counter to new observations ('the human was observed performing a negative behaviour, indicating that the human performs negative behaviours').
Contrition, repentance, or shame following negative or harmful actions also effects the perception of observers - 'I wish I hadn't done that' or 'I shouldn't have done that' rather than 'I wish you hadn't seen that' or 'I wish you weren't reacting to my actions in this way' indicates that the human being observed regards their behaviour as out-of-character for them, or out-of-character for the character that they are striving to achieve - and that information is relevant to how the behaviour should be weighed in expectations of the human's future actions. A human that responds to criticism with denial or anger will be regarded less favourably by observers than a human that responds to criticism with consideration or understanding.