By not putting reductionist labels on them.
For one, many studies of identical twins raised in separate households show they have the same personality type at a much higher rate than chance.
Two, there are incredibly strong correlations in the data. In different surveys of 100k+ people, the highest earning type has twice the salary of the lowest type. This is basically impossible by chance.
The letters (like ENTJ) correlate highly to the variables of Big 5, the personality system used by scientists. Its just that it's bucketed into 16 categories vs being 5 sliding scales.
Scientific studies are looking for variables that can be tracked over time reliably, so Big 5 is a better measure for that.
But for personal or organizational use, the category approach is a feature, not a bug. It is much more help as a mental toolkit than just getting a personality score on each of the 5 categories.
1) As I mentioned, it has a lot of statistically significant correlations, including to all the variables of the Big 5. Example: Surveys show that % of the overall population that is each type (like INFJ) is very consistent across time and populations.
2) Beyond that, youre right, there are a lot of personality systems with pros and cons. But Myers-Briggs has by far the must supporting materials, tools, ease of use, and so on. I think its the quickest to make useful to the average person.
3) I've found it really helpful as a lens for self analysis in my own life.
Anyway, their sibling comment told me what I wanted to know, so in that way I'm wasting more of my time contributing to this
But since nobody has mentioned the alternative yet, the framework used by anyone in any scientific capacity is the Big Five: https://en.wikipedia.org/wiki/Big_Five_personality_traits
The link between programming and conscientiousness seems fairly straightforward. To fully translate Mort/Elvis/Einstein into some kind of OCEAN vector would take a little more effort.