Not fully rational, but their average/typical behaviour is closer to the economical “rational agent” than our average is.

Let’s take your example of upvoting a "THIS 1000%” comment. For the rational agent, upvoting it has pros and cons:

• pros - allows the rational agent to express itself, support its own views, encourage the participation of people with a similar mindset.
• cons - makes it only slightly harder for the rational agent to find relevant info in the future; it also requires some effort (a single click).

If the pros > cons, it’ll upvote the comment. Otherwise, it won’t.

The upvote will also make it harder for other people to find relevant info, but the rational agent doesn’t care; not because it’s short-sighted, but because it’s 100% selfish. Its individual actions only make the platform slightly worse for itself, but they give it a really nice ego boost that more than compensates it.

And, if the platform becomes too shitty, the rational agent knows that it can simply migrate to another. But unlike us it’ll only do it once the new platform offers more subjective value than its former platform, by a small margin (what you called “inertia”, aka “cost of switch”).

Or in opposition to rational, there is maybe “emotional”, so that you have a feeling and want to express it, and you see something that expresses that, so you “like” it further, in addition to liking / upvoting the original comment - without considering the long-term ramifications.

The economy-style fully rational agent does take into account emotions. But only its own emotions, never the ones of other people.