LD asked for a thought experiment, not about the courts, perhaps to examine people's impressions prior to even getting into a jury, or to examine the impressions of the populace in consideration of voting.
Imagine there is a 50% chance each accuser is telling the truth and a 50% chance they are lying. Assume they are independent. What is the probability that all N are lying?
N Probability all are lying
1 50%
2 25%
3 12.5%
4 6.25%
5 3.125%
...and so on...
How confident do you as a person need to be to "believe" something? 51% confident? 80% confident? 90% confident? If you need to be 51% confident, then 2 people would be enough on average and with certain assumptions. If you feel you need to be 95% confident to believe something, then maybe that is 5 people for you to believe it.
As for me, I may end up believing something sometimes at less than a 90% confidence. However, I try to practice my life in a way where I suspend belief because it is a complicated thing based on intuitions and super-fast brain processes in some advanced algorithm that can be wrong. I would want to confirm my beliefs using evidence and arguments or suspend them in some cases when the outcomes have large risk for quick, rash decisions.
Also, importantly, the assumption of independence of accusers can be wrong. If we're talking about a particular accused person, like a rich or charismatic person, there may be multiple persons who all want to accuse them for the same reason. Or accusers could all be Republican operatives lying about an atheist, etc.
And the assumption of 50% telling the truth versus lying can be wrong and may be context based. So, for example, back in Salem in the 1600s, people lying about witches may have been common. Today, ladies lying about sex assault is rare.
Suppose we were talking about Salem witches then. Suppose accusations were independent [they often were not] and that this was about practice of witchcraft, not actually being a real magical witch. So, if you collected herbs and tried to poison someone, you may have been practicing but nothing magical about it. Now, suppose there was a 90% chance that accuser was lying. So we'd have:
N Probability all the accusers are lying
1 90%
2 81%
3 72.9%
4 65.6%
5 59%
So maybe you'd need 20 people telling you someone was a witch to believe they did some kind of Wiccan thing, like dancing in the forest or collecting herbs...but of course the magical claims would be 100% false at all times.
Now, suppose we are talking about women being sexually assaulted. Suppose accusations are independent [in cases of famous accused, they might not be]. Estimates range from 2% to 10% about women lying about such things. So, let's take the bigger number. Suppose there is a 90% chance that accuser is telling the truth. So we'd have:
N Probability all the accusers are lying
1 10%
2 1%
3 0.1%
4 0.01%
5 0.001%
That answer addresses the OP without any of the unnecessary assumptions about a trial or credability of a particular witness.
That answer highlights how the OP cannot be answered rationally, without extensive caveats and clarifications. You asked about some vague, variable, context-dependent state of "believe", which has no reliable relationship to probability estimates. Where exactly does the decrease in probability of all accusers lying = a change in "believe"?
Then, even by substituting probability estimates for "believe", Don's post has zero application to any real world situation, because it makes a logically impossible assumption that all accusations against a given person are independent. The "Independence" logically required by Don's method of computed combined probabilities doesn't just mean the accusers do not know each other, it means there is nothing that connects them in any way and no shared causal factors that impact the probability of each event. Thus, it is logically impossible for 2 accusers of the same person to ever be independent. Unlike coin flips, accusations (true and false) against a person are not random events. Who the accused is and many things about them has a causal impact on their probability of being accused, both falsely and truthfully. Which means those same factors impact the probability of each accusation against them, making each accusation against them non-independent from each other, by definition. Plus, countless things, such as how the accused is known to the accusers (e.g., are they famous and known to most people) impacts the degree of non-independence. The amount of non-independence inversely reduces the amount that the second accusation impacts the net probability. Since the amount of non-independence is always non-zero but its degree varies wildly by context, the hypothetical where total independence is assumed does not apply to any situation.
The only answer to the OP that has any applicability to actual world is something vague and qualified enough to actually be relevant to most situations, for example "
In general, as the number of accusers increase there will
tend to be some
variable amount of increase in the probability of the accusations being true."
Note that even there, "tend to" is a required caveat because it is possible for an added accusation to have no impact or even to decrease the probability of the accusation being true. Since the accusations against the same person are not independent, what is true of one accusation increases the probability that the same is true of other accusations. For example, physical evidence that the second accusation is true also increases the probability that the first accusation is true. Likewise, information that suggests that the second accusation is false decreases the probability that the first accusation is true. If two people make accuse a person and one of the accusations has direct evidence that it is false, then the probability that the other accusation is true is actually lower than if it was the only accusation made.