Well, I believe what you are asking does not have a particular "one-number-fits-it-all" solution.
What could possibly be done (even to an approximate degree) is to somehow think of the difference when the room is empty against the room being occupied by people.
The only difference of course is the people. So the main difference they will introduce to the sound is the absorption, which mainly affects reverberation. Of course there will be "spatial masking" for people standing/sitting behind other people but this is something you cannot solve by tweaking one volume knob.
I will try to "solve" a problem here to find what is the influence of the people to the reverberation time. So if we assume (for simplicity) that Sabine's formula is valid for your room we know that
So by subtracting one from the other to find the difference we have
We know of course that
and the absorption of the people can be found on some tables available online. So the total absorption of the people is this number (absorption for one person) times the amount of people. So, after some substitutions and manipulations we can find
where here A for people is the total absorption of all the people (what you found on the table times the amount of people present). You can see that the difference is the initial reverberation time scaled by the percentage of the absorption introduced by the people over the total final absorption.
Finally assuming (as we have already done in our choice to use Sabine's formula) that the field is diffuse we can approximate a halve of reverberation time with a halve in the total energy, leading to a decrease of 6dB. As can be seen from the above formula in order to achieve such a huge decrease in "volume" (abuse of terminology here!) we would have to introduce absorption equal to 2/3rds of the empty room's. Which is quite too much to ask from 20 people (in my opinion).
Now, all this nonsensical "derivations" have a lot of conceptual flaws (obviously), so first of all please do NOT consider this a mathematical or physically rigorous approximation. Now, the flaws are:
- The sound field is not diffuse.
- The absorption is not well distributed.
- The total absorption will be considerable.
- The whole room is not excited due to speaker directivity, which of course is frequency dependent.
- Possibly something else I can't think of at the moment.
Now, 1, 2 and 3 render Sabine's formula "not-so-useful". 1, also makes our assumption that halving the reverberation will halve the energy in some places in the room, incorrect.
With all those mentioned we can conclude that our estimate (whatever this will be), if based on what we did up to this point will be an overestimate (by a good amount) of the resulting reduction. So, as I said above, looking for a 6dB reduction even with this overestimated derivation is too much, even more when adding the fact that we have overestimated by a good amount.
To conclude this too lengthy answer (I strongly apologize for that, I just wanted to make my way of thinking as clear as possible to avoid misunderstandings) I believe that an increase of a couple of dBs will be quite enough (if even needed).