We can understand things much quickly when things are defined as a duality. Like good and bad, happy and sad, positive and negative, slow and fast etc. When two things are distinguishable distinctly the human mind tends to grasp it more easily versus defining it on continuum scale or having more than two states. Duality is the most easily understood phenomenon of human mind. That's why most classification happens on these lines despite the existence of a continuum of states.
There are many reasons why people tend to understand things in dualistic ways, but I'll mention two:
In the realm of inference (e.g., judging what is true, forming predictions), having mature schema for each end of a spectrum (e.g., one for introverts and one for extroverts) allows you to infer what positions between the two extremes might look like with adequate levels of accuracy. So, if you have developed schema for the ends of the extroversion spectrum, you can infer what a moderately extroverted person would be like without actually needing to store a third schema for a moderately extroverted person in memory. In other words, a dualistic understanding allows you to form schema for points between the extremes on the fly when you need it (sometimes called interpolation). It isn't a system that's free of the risk of error, but its easy to see how it is extremely efficient from a computational perspective in a world where agents don't need perfect mental models.
In the realm of decision-making, its important to consider the nature of what a decision actually is. While information and judgments might rest on spectra (e.g., the degree of attraction felt for a dating partner, strength of a preference for a political candidate) the actual decisions that are informed by these judgments are often discrete. For example, you can't "kinda" marry or "kinda" vote for someone. You either choose him/her or you don't. You can only express one behavior at a time. What this means is that information and judgments which fall on spectra often need to be converted into yes/no decisions. For example, when choosing which of five schools to attend for college, you're actually trying to convert a plethora of continuous information (e.g., benefits and costs of each school) into five yes/no values (specifically, one "yes" and four "nos").
This is such a ubiquitous phenomenon in decision-making that I don't think its surprising that people naturally tend to use binary language to speak about things which exist on continua. That said, its important to point out that people don't always reduce topics they care about to binary categories. In many cases, people will share their ambiguity or ambivalence with trusted friends when they are having a hard time converting continuous information into discrete decisions. Pay attention the next time a friend opens up about something difficult and you'll notice that their language often demonstrates nuance and awareness of grey areas.
The paper cited below is an excellent discussion of the relationship between continuous information and discrete judgments and the reference section is packed full of papers that go into much more detail than I'm able to accomplish here.
Luan, S., Schooler, L. J., & Gigerenzer, G. (2014). From perception to preference and on to inference: An approach–avoidance analysis of thresholds. Psychological Review, 121(3), 501. (link to PDF)