Reasoning is a stunningly complex topic -- how humans think. How can such a question ever be answered? Well, the approaches have been wildly different.
In cognitive psychology, researchers have focused on anthropological regularities, ways in which human reasoning is biased and deviates from the 'objectively correct'. Pioneered and popularized by Nobel-prize winning Kahneman as behavioral economics, the psychological work on reasoning seeks to explain how people reason answers to problems. The key point here is to focus on reasoning tasks that have an objectively correct answer, commonly probabiliities and set theoretical puzzles (famously, people judging that Linda is more likely to be a feminist bank clerk than to be merely a bank clerk).
In rhetoric, scholars (e.g. Toulmin's 1979 book, An Introduction to Reasoning) have looked at ways people reason in arguments, through discourse. In contrast to the kind of positivist studies of cognitive psychologists, this endeavor is largely descriptive. D.N. Walton has done enormous job in describing "argumentation schemes" -- ways through which people explicate their reasoning to one and other. Of course the study of rhetoric is also normative in a sense that some forms of arguments are fallacious and some are acceptable. Yet, a gray area exist: for example, slipperly slope arguments are commonly fallacious (unfair), but in some cases valid (whether they have a point seems to be genuinely an empirical rather than a philosophical question).
Both psychological and rhetoric work assume that proper reasoning is schematically correct and somehow independent of the content. This is a key assumption in producing generalizable scientific knowledge about how we think. Alternatively, we might conclude that this is simply not possible. This is what work on moral reasoning tends to do (Toulmin's 2001 book 'Return to reason' takes this approach). If reasoning is about fitting premises to valid conclusions, it might just be that the content of the beliefs we hold form premises that cannot be linked mechanistically to any predetermined set of conclusions. Moral principles can lead to a range of conclusions -- often because they spawn contradictions that escape unique resolutions. Religion can justify war and peace, and no form of 'moral calculus' will tell an artificial intelligence system what the 'correct' answer is.
This is a very disappointing view from the perspective of generalizable science. If our thinking is driven by rather idiosyncratic beliefs that can be used in a variety of ways not easily explained from a universalistic normative or descriptive viewpoint, there is only so much we can do with reasoning. I believe this is largely the case in areas well beyond ethics and morality. Managers and organizations more broadly are reasonable and rational, but there is no generic way to explain the uniquely most rational conclusion to be had from the commonly held/shared beliefs. Premises are connected to conclusions in ways that are neither arbitrary nor fully predictable based on some universal principles. Reasoning is messy, but not messy enough to be ignored.
These are my premises. There is a place and need for psychological study of 'biases' in human reasoning, but such biases are not the whole deal about reasoning. There is a need for a deeper understanding of generic rhetorical forms of reasoning and argumentation but these do not capture all that is relevant either. Reasoning is, to use some terrible jargon, more semantic than syntagmatic. This is the approach Foucault took in his genealogical work on rationalities and 'episteme'. I do not think his ambitious attemps that genelizable insight were very believable, but that is a topic for another posting...