The knowledge account of assertion (KA) says you may assert Q only if you know Q. KA’s defenders (e.g. Unger and Williamson) have frequently pointed to data on challenging assertions to support their view. For instance, when I assert Q, it’s appropriate ask ‘How do you know that?’ or ‘Do you really know that?’. KA nicely explains the relevance of these questions.
One objection to this is that it’s also appropriate to challenge an assertion by asking ‘Are you certain of that?’, even though knowledge doesn’t require certainty. KA cannot explain this. (It can if knowledge requires certainty, but this is a non-standard view.)
I have a KA-friendly story to tell about why the “certainty” challenge appears appropriate (which I won’t explain here). In addition to that, I also think KA’s proponents can and should expand the terrain of linguistic data, to take into account not just appropriate challenges to assertion, but appropriate prompts as well.
For instance, these two prompts are practically interchangeable: ‘What time does the meeting start?’, and ‘Do you know what time the meeting starts?’. We respond exactly the same way to either of these prompts, namely, either by saying what time the meeting starts, or by saying ‘I don’t know’. But these prompts aren’t interchangeable: ‘What time does the meeting start?’, and ‘Are you certain about what time the meeting starts?’. The latter is too demanding, and even somewhat alienating. Save for special circumstances, we don’t prompt assertion this way.
So whereas knowledge is closely connected to assertion both upstream and downstream, certainty is, at best, connected downstream only.
More on this and related matters here.