FreeOrion

I'm vetoing any system where diplomacy is built around a random chance for AIs to accept proposals from human players. This is bad for two reasons:
-It treats AI players significantly differently from human players.
-It requires AIs to to use a random chance to decide their actions. It should be possible to write an entirely deterministic AI that is fully capable of making decisions for itself, based on the gamestate, without resorting to dice rolls.

This is not the same as proposals that have two empires work out a treaty, and then submit that treaty to automated arbitration based on some measure of empires' or populations' opinions of eachother, or dipolomats' skills. I also don't like this, but it's not a veto on principle.

If this was intended for me, I am not suggesting a system where the AI would use a random change to decide its actions. I personally hope that we can create an AI that can make decisions based on the gamestate and is there for truly intelligent.

However if we are going to have complicated diplomatic treaties, it might be nice to offer the player some sort of a "diplomatic compass", which would help him/her to construct diplomatic treaty suggestions, which our intelligent AI is likely to accept. So for example you are suggesting the following kind of a treaty:

"We will form an alliance with you, if you offer us star systems X and Y."

When the player adds new components to the agreement the AI reads it and already makes a decision, based on the game state, whether to accept or to reject the treaty or make a counter offer.

So when the player adds the first component to the treaty, the alliance for example, the AI`s decision is yes. This is then presented to the player in a rough estimate, which for example might say: high change of success.

Then when the player adds the demand part to the treaty, the AI`s decision is no. This is again presented to the player in a rough estimate, which might say for example: low change of success.

This way the player has some sort of an idea of whether the treaty that he/she is proposing is reasonable or not, before he/she actually proposes it. This is particularly important if we are going to have complex treaties, so that we could avoid most of the frustrating treaty reshaping that eleazar was talking about. However if the information that is presented to the player is always absolutely correct, when evaluating changes of success, there is no excitement in the diplomacy system, since you already know before hand that the treaty that you are going to propose will be accepted by the AI.

_________________
What is your favourite alien species and WHY?
Preliminary thoughts about diplomacy
Some unfinished ideas for specials

Goodmorning all,




Fair enough, My hope is that such a Diplomat methodology could Remove the NEED to tinker and play, or guess weather a treaty will go through, or even care if what your presenting will be accepted.

Instead, you simply write your dream treaty

"They give me some food each turn *without specifying how much*, they give me these three techs, they give me *a, a few, some, many, lots of, all their* ships, and two planets *by name*" *in that order of importants*

The other party(ies) would do the same thing.

Before sending in Diplomats each leader would get the dream list of each other leader, and re-order them in acceptability. (this would be pre-initiated to something sane, so the user doesn't HAVE TO do this.)

in my example, the other teams dream aggreement might be

"They give me pp each turn, they agree to a mutual research plan for tech B[of the type Krikkitone proposed, 100 RP by me is 100 RP for me 30 for him ]' (this player is probably AI, so by programming doesn't include Stupid Requests).


Clause lists can update in real time, or semi-real time, because it's not so important what you put on the wish list.

Second phase (can be done in parallel just separating into phases for describing):
The AI looks at my list, and I look at theirs, or a fellow humans.

now the AI is easy, They AI is going to move my last three request to the Vedo list, and indicate that the food is acceptable in principle. (These would be simple things, a veto list or a red X, a green check for 'ok as is', green up arrow for 'I want this or more as well', and a yellow down arrow for 'our diplomats have low exact numbers/less powerful clauses in mind, but this idea is ok')

On my side,

I don't have many pp's so i might give the pp trade a red X or move the AI's research treaty above their pp trade, and, if more serious about not giving them up, give the pp trade a nice yellow down arrow *moving above would indicate more willing to give then anything below* {hopefully the game would be able to look at my PP situation and minimum pre-rearrange the two demands so that pp's is lower because I'm less likely to be willing to give them.}

then we would launch the Diplomats.

The Diplomats would then compare the amount of food Empire B has spare /turn, the amount of pp/s i have per turn, and (perhaps if we want) the current going trade rates of these, and weather or not either team put green or red arrows to set Abstract 'value/commodity' values for each side, In a sense converting each offer into a price, then it would convert the research clause into the same value scale, based on the relative size of each teams research budget.

Then it would calculated any trade bias, *based on Diplomacy choises at race pic time, populations opinions from each side*, special buildings, all the various factors which could make one of the two empires 'profit' more from an agreement then the other.

After the trade bias is worked out, it would adjust the free variables how much food/pp/research% per turn, until equilibrium is reached.


Now all empires would be presented the conclusions. With Number filled in, in place of abstracts.
"Some Food"/turn becomes 13 food/ turn, And the biggest reason why any clause was left out *the snaglres veto'ed your request for* *probably simply an x and who from*.

you would look at the results, and accept/rearrange priorities and rerun, or refuse to sign*without penalty*.

If i had put at the top of my list, something that they Veto'ed there would be an additional chance that talks would fail.


So quickly I'd like to touch on WHY this system is good:

One, I think we want Diplomacy ability to be a race pick able trait. As such natural consequences of high and low diplomacy skill is higher/lower probability of accepting clauses you propose, and higher/lower returns for trade.

If we want a trade system of X food, Y PP/ per turn in exchange for O money, and P research %/turn. we need SOME method of re-including those race benefits.

if we don't use *some* form of algarhythmic processor to ballance the trade automatically and take into account ratial pics, population happyness, yadda yadda. . . the only option i see is having all incoming trade multiplied by some set amount. *Trade Wizzes, All incoming money via trade is increased by 15%* which seams far too basic and unfun for FO.

Also, let's say you've got 10 food/turn you don't need. Either you can go to each group, and make demands and offer your 10 food, and play around with demands to see what each will give you, then half an hour later, choose which group gave you the best offer . . .

or you can open up the trade window to all your neighbors (no reason not to do so in parallel), Input what you want back, pp and money, and what your willing to give food, or 10 food , and have the diplomats return the various possible trades with various empires including the rates they are willing to give you of each pp and money for your 10 food, and the political consequences of each treaty. And in the end you chose that which wet your whistle the most.

Two; With the proposed system you know your always getting the best possible deal you can get, never any NEED to relaunch the trade request asking for 1 more food to see if they accept it, unless you fundamentally change the incoming parameters *add remove clauses, veto.unveto clauses, or move clauses* the results will remain constant and optimized.

Three; Human to Human trade managed in the same way as Human to AI trade. *Moo2, Repulsive species COULDN'T trade even human to human. Seams a bit binary that system, compared to FO's metered systems*

four: super versatile, if well made intuitive, and powerful, but with minimal micromanagement required and minimal benefit for micromanaging.

I'm sure i could ramble more, but i'll stop.

Best wishes, Robbie

In terms of a Diplomatic Review, what would be simplest is to have the AI propose several acceptable counteroffers to any of your offers.

In civ IV they do this well, you send the AI a Diplomatic proposal and say "What will you give me for this?" ie ask the AI to put more stuff on Its end. OR you ask "What do you want for this?" ie ask the AI to add stuff to your end that it wants from you.
(they can respond with sorry, ie they/you don't have enough to make a deal)

Generally in civ IV there are limited things that can be added which is why in FO it might be good to have the AI come back with a few Counter offers (all of which the AI would accept)

Now admittedly that gives the Human an advantage over the AI in terms of Haggling (the Other Player might be willing to pay 100 gold but you would take 50 as an offer. If the Other player is an AI then they will tell you the 100)

I'm playing around with the idea of improving on this but I'm not sure if it is possible because that would be a solution to a 2 party Bartering problem which would be probably impossible to do in an Algorithmic way that was clear and not easy to manipulate.

The key thing for the Bartering to work with an AI in a non manipulating way, the way it works with a human, is that what the AI will accept depends on what Treaties you've already proposed.

What MIGHT be the best is
STEP 1. You construct a Treaty with the help of a 'diplomat' ie

I say I am willing to give up Fusion Beams, what will Player X PROBABLY give for that?
I say I want Mars, what will Player X PROBABLY want for it?

STEP 2. your 'diplomat' constructs a few treaties from thos requirements with one of 2 characteristics
1. Highest chance of success (ie at this point the Treaty looks like it is a sure thing, according to 'your diplomat' the other player will Jump at the chance
2. Best Return (ie this seems like the absolute most you would Ever get out of the other player... according to 'your diplomat' at this point the other player would be just breaking even)

'Your diplomat' can also rate any proposed Treaties in terms of how 'good a deal' it SEEMS to be for the other player.

STEP 3. You send the Treaty to the other player, at the end of the turn they accept or reject it. If they accept it, it becomes an active Treaty. If they Reject it, then you (or they) can make another offer Next Turn.

[once you send a treaty to another player you have 1. agreed to follow it yourself and 2. given it as evidence of something YOU will accept, so if you are willing to give 100,000 gold for a planet... then the other player knows you want the planet a lot and they might not take the deal even though they would Normally only need 10,000 gold for the planet]


A key thing here is that 'Diplomat' to make the diplomat useful, it should calculate the 'value' of a Treaty based on MORE information than you have. Ie You may Not know (through spies, etc.)that the other player is low on food, but if your Diplomacy Rating is high (particularly with respect to the other player) than your 'Diplomat' will tend to construct Treaties that put a really high value on food. (at a minimum your diplomat will construct Treaties based on the Info you Know).

Now there is some info the 'Diplomat' obviously cannot Know like the strategy of a player (human or AI) but they should be able to guess at it from the game state. You may know that your Friend NEVER gives up a planet, so the diplomat's suggestions aren't going to be useful. But if you are playing against an AI or against unknown Human players the 'Diplomat' should be a decent guess of how they are.

This means there is only one round of 'Negotiation' per turn.. but your 'Diplomat' acts as sort of a 'Pre negotiation partner' so proposals can go back and forth multiple times between you and your Diplomat, but then are only sent out Once to the Other Player for 'Actual Negotiation'.

(Because Actual Negotiation with an AI Player is bad one of 2 ways
1. You can get the best deal possible by starting at something unrealisticlly good for you and making it slightly worse step by step until you get to the first Deal the AI player will accept
OR
2. To avoid this, the AI player modifies its valuation of deals based on deals you've proposed, but that prevents the player from exploring the options of what the AI wants to try and get the best deal (it makes it all complete guesswork).

The essential Ideas here
1. Acceptance/Rejection of Treaty is done by the Player in their own way which is UNKNOWN to the other Player (in a human way for human players, and deterministic for AI players but the AI's deterministic workings are not known to the other players... because it depends on parts of the game state the other players don't know)

2. The "Negotiation" of a Treaty is done through a 'Diplomat' which is algorithmic and allows you to design a Treaty the Other player will Probably Accept (based on the game state including which Treaties they have Proposed/Accepted/Rejected in the past) The amount of the game state available to the 'Diplomat' **could** be increased or decreased [with a minimum of the 'game state' visible to the player... but extending up to all aspects of the game state knowable to any player.]

** This is NOT Necessary, Diplomatic picks/techs/effects can work far more easily through affect the "Citizen AI"

Edit by Geoff: Moved "Citizen AI" stuff to Simulating Citizens thread.

The server is not going to send information to players that they don't already know, and no plans or systems that are based on this are going to be accepted. If such information is needed for a diplomat AI or other purposes, players can get it using traditional recon or espionage.

Suggestions of diplomat AIs in general aren't likely to happen in v1.0. Even if they did, they shouldn't form the basis for the diplomacy system, such that we assume they will be there and design based on that assumption. This is vaguely similar to the economic system, in which we avoided using an overly complicated system that required an AI to play the game for you. Likewise for diplomacy, we can't rely on an AI to make treaties for the player.

We *can* build in an offer-counteroffer system, however. The interface for doing this would be the same whether playing against AIs or other players. After setting up a treaty offer, you'd send it, and a message would be sent to the other player containing the offer. For humans, this would display the message and enable the player to accept, reject or alter the offer and send back. For AIs, they'd get the message, analyze the offer and respond accordingly. The human version of this could be presented as an integrated GUI with offer and counter-offers displayed with treaty messages shown immediately and repeatedly until a deal is reached or rejected. Alternatively, the messages could be treated like chat messages "Player X has offered a treaty... click here to view / respond", which the player would have to open to see the details, and could ignore or respond whenever they want to. (AIs would be programmed to respond immediately to any treaty messages in either case)

Doing this would require predefining all the possible treaties we'd allow in the game mechanics themselves. AI's wouldn't be able to be written to have new / different types of treaties with clauses that the game engine doesn't provide. Human players would be able to avoid this limitation by additional negotiation with plain text, however, which is unavoidable and not really a problem.

_________________
—• Read this First before posting Game Design Ideas!
—• Design Philosophy
—•— My Ideas, Organized —•— Get an Avatar —•— Acronyms —•—

I like the idea of offers and counter offers, but I'd like further clarification on what you mean by:

Dose that mean you can't have a clause based treaty builder like I originally proposed? And if so why not?

This doesn't say whether or not treaties can / should be editable and complicated or only prefab and simple. It just says that whatever clauses (specific points agreed to / in the treaty) that are possible to have in a treaty will need to be predetermined and coded into the game engine itself. The alternative would have been to let AI scripts somehow define whatever treaties are wanted, as complicated as the script writing can manage.

Well things I really want to avoid in Diplomacy in terms of how the AI will do it (in priority)
1. Micromanagement (see civ 3) you narrowed in on the best deal the AI would give by making proposals asking "is this OK" and if it was you upped the proposal on their side or lowered it o... this also can be a problem working in the opposite direction, start high and slowly lower it until you got the offer they accepted)

2. Penalty for trying out ideas... the reason you can't manipulate a human player like that is that they will change their "price" based on what they think you will give, and plus they will get annoyed if you keep sending ridiculous offers. Now when I am playing single-player I don't want to have to look at an Excel spreadsheet to calculate what a Reasonable offer would be, so I need some ability to try out Ideas.

3. Exploiting the AI player (see civ 4) in civ 4 you made a proposal and the AI automatically told you what was the best deal it would accept with regards to that (or it said that it couldn't make a deal) This was better in that there was no Micromanagement, but it provided vast ability to manipulate the AI (The AI doesn't want to go to war... oh they are already planning on going to war with someone...I better prepare)




I could see revising my proposal like this

1. The AI player looks at FOUR things in any agreement
What AI gets - What AI gives [Must be positive for a deal to go through]
V.
What the OP gets - What the OP gives [the lower this is compared to above the 'better the deal' is for the AI... the more likely they are to accept]

ie when I sell a Tech, I don't 'give' anything... if I give a tech away my empire is no worse than it was before... but your empire is better, so I need something to Balance out the Net Benefit to you. (ie a Tech worth 100 diplomacy points 'should' sell for transferred goods worth 50 diplomacy points I get +50-0 and they get +100-50) The whole reason we would have trade in diplomacy is that I want X more than you do.

2. The AI player looks at what your Previous "Real Offers" have been in Guessing how much you value something (because of this only "Real offers" ie treaties that I would be bound by, can get 'sent' to another player Human or AI.. otherwise the AI is at a disadvantage because it doesn't consider 'query offers')
IF I start by offering the AI a really bad treaty, (give me your empire for a dollar) then the AI should be less willing to give me its best deal (and it will tend to give me similarly bad offers)

3. You have some type of a guide 'diplomat' that looks at all the information you know [including previous proposals sent] and helps design a proposal that it thinks will be accepted (Essentially the same thing the AI player would use in designing a Treaty to get sent to you, but they have different info.)

The diplomat would Not be key to the design of the system, but if my sending of an offer is going to be way too over or undervalued, I Must have some way of knowing. Especially if my sending an over/undervalued offer to the AI changes how it responds to my future offers (which it must to be a 'non manipulated' AI).



As for the predefined Treaty terms by breaking the down to simple components, it wll work well

[one point, one sided Peace should probably Not be an option... as it would be an Incredibly stupid deal to take]

One thing I would like to see is if Multi-Party Treaties can be done
Ie a list for each empire of what is given and what is gotten

Red gets
Peace from Blue
Peace from Yellow (25 turns)
Fusion power tech [doesn't matter from who as long as someone in this Treaty has it]

Blue gets
Peace from Red
Defensive Agreement from Yellow (50 turns)
Peace from Yellow
Mars (from whoever has it)

Yellow gets
Peace from Red (25 turns)
Peace from Blue
100 Minerals from Blue
50 Money from Red
6 PP/turn from Blue
4 Minerals/turn from Red (10 turns)


Yellow proposes it and both Blue and Red must accept it.



On Breaking Treaties, this gets complicated when you have an 'ongoing' treaty especially if a one time thing was also part of that treaty

we should probably have
1. the ability to suggest a renegotiated treaty without first breaking it (ie suggest that instead of 10 food/turn and Peace in exchange for Peace it becomes 5 Minerals/turn +Peace for 5 PP/turn+Peace) If Both sides agree the change is made with no need to Break the Treaty [although the Threat of breaking the Treaty is there]

2. If the Treaty is actually Broken, the "Citizen Response" should depend on what the Citizens thought of the Treaty (if the population of a Planet though that treaty was 'unfair', it might be a net benefit to you on that planet)

Oh that's fine , I'd always envisaged a hardcoded list of possible treaty components, a long list but a list nonetheless

I don't see why so many people want some kind of AI diplomat. It's totally redundant because we have a real player making the decisions and he should be making his treaties based on his interpretations of the other empire's position rather than that of some AI diplomat.

_________________
Warning: Antarans in dimensional portal are closer than they appear.

I don't want to muddy the waters here. But a potentially significant subset of "diplomacy" is the "First Contact".

There's a thread on that here.
My vision of a relatively simply, but strategically significant first contact is this post.

_________________
—• Read this First before posting Game Design Ideas!
—• Design Philosophy
—•— My Ideas, Organized —•— Get an Avatar —•— Acronyms —•—

Simple... I'm a new player to FO, I have a Tech I want to trade, How do I know how much I Should ask for it?

If the AI is going to react to significantly bad/good offers by changing what they think MY position is, at least I should have SOME idea what a good deal is.
Because if we have a truly "Diplomatically Competent" Player AI then this situation will come up
If I offer the Tech for 100 $, I would have gotten it
BUT
Because I offered the Tech for 250 $, I can't sell it for 100 $ now (they figured I was playing hardball, when really I really just had noIdea about how to value a Tech in terms of $)
OR
I offered it for 20 $ but had no idea the AI (or any decent Human player) would be willing to pay 100 $

The point is Guessing games/micromanagement is a bad thing. While we could simplify the Economic system so that the Guessing games would be unnecessary, we can't simplify the Diplomatic system beyond a certain point because Players are a key part of the Diplomatic System.. one that we can't simplify.

So we either
1. Simplify the AI players (making them easier to manipulate than Human players, and almost Totally predictable diplomatically)
OR
2. Give the player some assistance in dealing with the AI Players. (and if the AI Players are any good+the Assistance is any good, that will be useful for doing the Human Players)

..........................................
On "First contact" that is part of what I was thinking of with the 'Level of Relationship" at First contact it is 0 (or 1 out of 100 or something)... and your 'level of contact' could be a race Pick (ie do you have records from the time of the Orions so you know about other races, and can start First contact at a higher Level, or are you reaching into space for the first time, and the Orions ignored you)

That's a big assumption. My guess is that it would take an extremely sophisticated AI to make useful reactions based on that input while avoiding being easily manipulated by a human.

Do you know of a game that claims the AI reacts to that kind of info?

EDIT: to clarify: this post is based on the understanding that,
"the AI is going to react to significantly bad/good offers by changing what they think MY position is.",
means something like,
"The AI is going to make assumptions about the overall strengths/weaknesses of an empire based how it bargains, and apply those assumptions to it's overall strategy."

_________________
—• Read this First before posting Game Design Ideas!
—• Design Philosophy
—•— My Ideas, Organized —•— Get an Avatar —•— Acronyms —•—

You don't. Learning that is part of learning the game... But more so, it's not well-defined number; each AI script in each situation in-game will place a (wildly) different value on any tech, and it's not possible to have a single useful valuation estimate for all such situations, even if the estimates attempts to takes into account what it thinks should modify the value. Also, the in-game market is rather illiquid - there's no really a lot of alternative sellers and buyers for any in-game commodity, so the price is really whatever someone's willing to pay, and the cost is whatever you can get someone to sell it to you for, and neither is solely a function of the in-game situation or market consensus.


What's wrong with asking the AI "What would you give me for this?" as in Civ III ? There's no difference in terms of the mechanics of haggling (or arranging things to avoid haggling if so desired) if one player or the other makes the first offer.