Growing up, I've always placed psychology as a category separate from all the "hard science," some call it "soft" science but as cheesy as it sounds, I'd rather call it future science.
Artificial Intelligence is a bot or android that can understand human emotions; it can communicate on a level that a typical human can.
So, technically they already used the basic principals of psychology to interpret the way a human responds and convert that into their own mechanism of response.
Psychologists are experts on such activity, they are trained to think like that, they put on the masks of their subject and solve a puzzle to unlock the secrets of their subject's next moves.
The scary thing is A.I are novices in this technique in the beginning but eventually, over time, they would become experts in this field, they would be quick to recognize any human instinct and read their mind.
This is part of what makes Artificial Intelligence intellectual.
Another method they use is the mirror learning, mirror learning is how an android receives and understand its surroundings.
The topic of empathy is always considered something that an android cannot possess, they don't have the ability to understand different human emotions on the same level that other fellow Homo sapiens can.
This fact is displayed in the book Do Androids Dream Of Electric Sheep? where Rachel, a Nexus-6 android, is programmed to coax and have sex with the hunters who try to hunt her.
This was something that happened to a character in the novel who developed feelings for the android.
Is what Rachel did completely insensitive? No, it wasn't, she isn't a human being, throughout the book, some readers would forget the fact that she is an android because of the way she was seen by the author.
She was written as "human" as possible in the novel, her empathy towards other androids was shown on the empathy box.
This could've tricked the readers into thinking that Rachel was an android with emotions because empathy is an important element in being human.
She was programmed to be that way, and she would try whatever she can to continue and succeed in being that way.
Let's take a scenario that androids were part of our government back in the 1960s and 1970s when America was in a diplomatic controversy with the Soviet Union.
If a secret organization from any two countries would program an android, then run a simulation where the android is in the position of the government and making important decisions, when the android is programmed to destroy their enemy, the simulation may result in the other country being completely bombarded with nuclear weapons without concerns for their own nation.
There was something called the Mutual Assured Destruction (MAD) back then, which was if USSR launches a nuclear weapon, the Americans can detect it before it hits and they could launch their own nuclear weapon.
Then it is MAD because most sides would have catastrophic effects.
The android was programmed simply to assure the destruction of their enemies without any concerns of their own safety because they only cared about the success, not the reverse consequences.
This is why the intelligence of an A.I android depends on the way they are programmed.
How the minuscule bits of details could prove to be extremely important to the way an android would think, if any "thing" is left out, there may be a chance they would be able to understand that part of existing.
In conclusion, the intelligence of an A.I depends on the mechanics of its programme and the intelligence of its programmer.