Post by chriscrawford on May 3, 2016 10:24:03 GMT -8
I have been thinking hard about how to implement lies in the game. These are extremely important to the game, so they must be done well. There are many factors to consider in telling a lie:
First, there are four different situations in which lies are possible:
1. I have been asked to reveal an auragon count.
2. I wish to propose a deal for an auragon count.
3. I am gossiping about somebody.
4. I am telling news about somebody.
The simplest of these is #1. Here I have already had three factors imposed upon me:
a. To whom I will lie.
b. About whom I will lie.
c. About what I will lie.
In this case, the only questions are whether I:
a. tell the truth
b. overstate the auragon count
c. understate the auragon count
d. overstate my certainty
e. understate my certainty
The criterion for the decision is the strategic benefit of my action minus the risk of being caught. Let's call them SB and RC. The equation looks like this:
Merit = SB - RC
But wait! The risk of being caught is not itself the bad thing; the bad thing is twofold: the loss of trust and the significance of that loss of trust. If Subotai doesn't much trust me anyway, I can lie without harm. Of course, if Subotai doesn't trust me, then there's really no point in making a deal with her, is there? Perhaps the real damage will be to my reputation with others, if Subotai tells them that I lied (and they believe her). Then there's the matter of when I face this decision. A lie early in the game will have effects that last through the entire game; the loss of reputation late in the game will not be so damaging. Hoo, boy, this is going to be difficult. Nevertheless, I think that I'll confine the calculation to risk of being caught scaled down by the amount of time remaining in the game. Hence:
Merit = SB - RC * (amount of time left in the game AT)
Now, what strategic benefit do I get from lying? After much contemplation, I have decided not to engage in complex calculations based on how many auragons I think that other people have. The only consideration should be my desire to harm my interlocutor. This in turn is based on two factors: my esteem for them (esteem = pGood + pHonest + pPowerful) and my fear that they're close to winning.
But wait! Should I add a consideration for how much of a threat I consider them to be? For example, Camiggdo starts the test game without any blue auragons. That makes her vulnerable to anybody who has a lot of red auragons. Should she concentrate on bringing down that character? Actually, she could accomplish this by telling the truth. So here we have another strategy: tell the truth to bring down a powerful opponent. I win in two ways: my worst enemy is laid low, and my interlocutor trusts me more.
This all seems entirely too messy, especially because I don't control what the other person will ask for. I could introduce a new verb "counteroffer", that allows me to offer a different value, one that I *want* to tell about, but again, that's really messy. So let's just stick with an overall assessment of threat in order to decide whether to lie.
There *is* one way to measure overall threat: adding up the combinations of their pAuragonCounts versus my corresponding AuragonCounts. For example, Camiggdo starts with 3 Red, 3 Green, and 0 Blue. Suppose that her pValues for Subotai are 1.8 Red, 2.6 Green, and 1.1 Blue. Her threat to me is then calculated like so:
1.8 SR * 2 CG + 2.6 SG * 0 CB + 1.1 SB * 3 CR = 6.9
meanwhile, if Zubi has 2.2 Red, 1.8 Green, and 2.3 Blue, then her threat to me is:
2.2 ZR * 2CG + 1.8 ZG * 0 CB + 2.3 ZB * 3 CR = 11.3
OK, I can use this calculation to determine who is the greatest threat to me; they're the one I want most to hurt. How about this general rule: Lie only about the greatest threat.
I'll go implement some of these ideas to get started.
First, there are four different situations in which lies are possible:
1. I have been asked to reveal an auragon count.
2. I wish to propose a deal for an auragon count.
3. I am gossiping about somebody.
4. I am telling news about somebody.
The simplest of these is #1. Here I have already had three factors imposed upon me:
a. To whom I will lie.
b. About whom I will lie.
c. About what I will lie.
In this case, the only questions are whether I:
a. tell the truth
b. overstate the auragon count
c. understate the auragon count
d. overstate my certainty
e. understate my certainty
The criterion for the decision is the strategic benefit of my action minus the risk of being caught. Let's call them SB and RC. The equation looks like this:
Merit = SB - RC
But wait! The risk of being caught is not itself the bad thing; the bad thing is twofold: the loss of trust and the significance of that loss of trust. If Subotai doesn't much trust me anyway, I can lie without harm. Of course, if Subotai doesn't trust me, then there's really no point in making a deal with her, is there? Perhaps the real damage will be to my reputation with others, if Subotai tells them that I lied (and they believe her). Then there's the matter of when I face this decision. A lie early in the game will have effects that last through the entire game; the loss of reputation late in the game will not be so damaging. Hoo, boy, this is going to be difficult. Nevertheless, I think that I'll confine the calculation to risk of being caught scaled down by the amount of time remaining in the game. Hence:
Merit = SB - RC * (amount of time left in the game AT)
Now, what strategic benefit do I get from lying? After much contemplation, I have decided not to engage in complex calculations based on how many auragons I think that other people have. The only consideration should be my desire to harm my interlocutor. This in turn is based on two factors: my esteem for them (esteem = pGood + pHonest + pPowerful) and my fear that they're close to winning.
But wait! Should I add a consideration for how much of a threat I consider them to be? For example, Camiggdo starts the test game without any blue auragons. That makes her vulnerable to anybody who has a lot of red auragons. Should she concentrate on bringing down that character? Actually, she could accomplish this by telling the truth. So here we have another strategy: tell the truth to bring down a powerful opponent. I win in two ways: my worst enemy is laid low, and my interlocutor trusts me more.
This all seems entirely too messy, especially because I don't control what the other person will ask for. I could introduce a new verb "counteroffer", that allows me to offer a different value, one that I *want* to tell about, but again, that's really messy. So let's just stick with an overall assessment of threat in order to decide whether to lie.
There *is* one way to measure overall threat: adding up the combinations of their pAuragonCounts versus my corresponding AuragonCounts. For example, Camiggdo starts with 3 Red, 3 Green, and 0 Blue. Suppose that her pValues for Subotai are 1.8 Red, 2.6 Green, and 1.1 Blue. Her threat to me is then calculated like so:
1.8 SR * 2 CG + 2.6 SG * 0 CB + 1.1 SB * 3 CR = 6.9
meanwhile, if Zubi has 2.2 Red, 1.8 Green, and 2.3 Blue, then her threat to me is:
2.2 ZR * 2CG + 1.8 ZG * 0 CB + 2.3 ZB * 3 CR = 11.3
OK, I can use this calculation to determine who is the greatest threat to me; they're the one I want most to hurt. How about this general rule: Lie only about the greatest threat.
I'll go implement some of these ideas to get started.