There's a tacit assumption that the discrepancy in emphasis and word choice is based on gender stereotypes, completely ignoring the possibility that men actually tend to be more aggressive and women actually tend to be more communal.

The statement "Subtle gender discrimination continues to be rampant," is completely unsupported by the evidence.

If we accept the common claim that women are discouraged from being assertive and behaving in "male" ways, then wouldn't we expect to see an behavioral difference? If there is a behavioral difference, wouldn't we expect that to be reflected in letters of recommendation?

This is shoddy science (failure to control for confounding factors) and implies shoddy social policy (maybe we shouldn't use letters of recommendation). What makes it worse is that it ignores what would be actual useful questions like the study that asked people to describe a video taped baby's behavior with some told the baby was a boy and some told it was a girl.

Or, you know, women might on average actually be more kind, communal and tactful hence why their recommendation include these terms...

Men might on average also be more ambitious and outspoken.

This could be attributed to research suggesting women value interpersonal relationships more than men and that men take greater risks than women (hence terms like outspoken, assertive and daring).

I have a good record on recommendation letters for grad school candidates always centered on stories of how a candidate has an insatiable thirst for the truth and challenges the status quo to satisfy it. I also make it a point on how this improves debate and understanding within their team improving on overall dynamics. Of course this is in the realm of high-tech engineering.

I think that probably does translate to assertive qualities, but in a sense this is what grad school is about, the search for truth and how a candidate will improve the experience for their fellow students and the faculty...

The important thing in effective recommendation letters and applications in general is to know what the evaluators are looking for and speak to it. If they are looking for communal cultural fit, by all means speak to it, but if not, it is a wasted opportunity.

I think it's telling that I knew subconsciously the instant I saw the 'communal' list that I would never, ever want a colleague or superior to describe me that way, while I'd be happy to have any of the 'agentic' list applied to me.