The field is over-crowded, for one. Two, its participant pool is almost exclusively undergraduate students on university campuses in WEIRD (western, educated, industrialized, rich, and democratic) nations. Three, experimental designs may rely on questionably reliable, possibly retrospective self-report to infer patterns of affect and cognition. Four, experiments are often conducted in highly contrived laboratory environments. (In-situ data collection as people go about their lives is becoming more common with the rise of smartphones and wearables, however.)
As in most domains of science, researchers are also pressured to publish regardless of the quality of the work, and replicated findings are usually not considered interesting or valuable enough to publish.
This is anecdata, so take it with a grain of salt. (I minored in psychology and worked as a research assistant with a social/cognitive psych lab as an undergrad.) Also, there is a ton of extremely good science in psychology. It's a shame the recent profusion of low-quality work obscures it.
As the other commenter mentioned, cognitive psychology is well-developed (working memory, the relationship between action impulses and our subjective awareness thereof, etc). And there is very good social psychological research on a number of useful topics including diffusion of responsibility, the fundamental attribution error, the effect of dissenting voices on groupthink, the influence of perceived authority on human decision-making (Milgram), etc. Kahneman and Tversky's contributions (and all of "behavioral economics") could also be considered social psychology.
Anyways, though, I'm not sure it's super useful to "believe with conviction" in science. Shouldn't we hold all results up to scrutiny, and weigh them on the evidence?
Along these lines, by far the most important takeaway from the last 50 years of academic psychology, imho, is that we are usually far too willing to trust our own personal judgment. The brain is as much a self-deception engine as it is a reasoning machine. We have a far hazier view of what actually goes on in our minds than we usually think we do.
Cognitive psychology topics such as memory and reading have a strong paradigm, and established results. In general, their effects are much easier to replicate because you require far fewer participants. This is because you can validly make multiple observations of the same participant.
As in most domains of science, researchers are also pressured to publish regardless of the quality of the work, and replicated findings are usually not considered interesting or valuable enough to publish.
This is anecdata, so take it with a grain of salt. (I minored in psychology and worked as a research assistant with a social/cognitive psych lab as an undergrad.) Also, there is a ton of extremely good science in psychology. It's a shame the recent profusion of low-quality work obscures it.