Static Actor–Motivation Forecasting: Peace in Middle East
We will do this backwards. Before I write the article that explains the model, I will give an example of how it is used. Why? Because I want to be able to make last-minute tweaks to the code as I write this article, and then be able to publish the “finished” code in the next article.1 In reality, of course, it won’t be finished. I will keep tweaking it as I use it.
What’s in this article will be my best impression of what I’ve learned from reading a bunch of Bueno de Mesquita. This will be a deepening of the steady-state forecast we discovered first using a refinement of the early War Trap model we also saw recently. (You may benefit from reading those articles first.) We will still not go fully dynamic today, focusing instead on baby steps.
Where we left off, we had just asked
Will there be a peace agreement in Gaza?
Let’s find out.
But first, I should reiterate that I know next to nothing about politics and international conflict. Please excuse my mistaken judgments in this article, and if you find something particularly egregious, email me and I’ll issue a correction.
Our actors once more
We had a set of actors in the steady-state article, but I’ve since read the news – possibly getting more knowledgeable in the process – and refined the set of actors to the following:
- The Israeli government (
IG), which wants a continuation of the current level of hostilities; - The part of the Israeli people who dislike the war (
Idfor doves), and those in favour of it (Ihfor hawks); - Hamas (
HA), who also thinks the current situation is desirable; - The Palestinian Authority (
PA), who would prefer a full peace as proof that Arabs and Israelis can co-exist; - The us (
US), who sorta-kinda want a ceasefire but not enough to cut off Israel’s supply; - The Iranian government (
IR), who wants increased hostilities because it furthers their position as a local hegemon; and finally - We will be lumping together some other regional powers like Egypt, Qatar,
Jordan, Saudi Arabia into a single peace brokering force (
PB), because they appear to have a similar desire for a ceasefire to begin with.
We enter these actors into the model:
model = AMModel([ # Sal. Inf. Pref. Actor("IG", 1.0, 0.7, 0.2), Actor("Id", 0.9, 0.2, 1.0), Actor("Ih", 0.9, 0.2, 0.2), Actor("HA", 1.0, 0.1, 0.2), Actor("PA", 0.9, 0.3, 1.0), Actor("US", 0.8, 0.9, 0.5), Actor("IR", 1.0, 0.9, 0.0), Actor("PB", 0.8, 0.6, 0.5), ])
Here, we leave out some other obvious actors that have a stake in the dispute.2 The people in Gaza, various agreeing members of the un, Houthis, Hezbollah, and China are some that come to mind. The reason we leave them out is that in my estimation, their power levels for this dispute are low enough that they are minor players in deciding the near-term prospect of peace in Gaza. Including them would not improve the predictive power of the model, it would just clutter the display.
Recall that on this scale, 0.0 means intensified hostilities, 0.2 means continued hostilities, 0.5 means ceasefire, and toward 1.0 would be a full peace agreement.3 After publishing this article, I’ve had some regrets regarding how this scale is constructed. If I did it over, I would likely use a different scale, but I don’t want to re-write the article now.
It may be surprising that an actor like Iran seems to wield so much influence.4 I certainly gave them less influence in the previous article. The reason is not that they are very convincing at a negotiating table, but rather that they are good at fueling conflicts in the region through intermediaries. They can make regional countries act a certain way not by smooth talking, but through use of proxy force.
At this point, we can look at our new power landscape.
print_landscape(model)
0.0 ######## IR 0.1 0.2 ######## IG Ih HA 0.3 0.4 0.5 ########## US PB 0.6 0.7 0.8 0.9 1.0 #### Id PA
We note that since the last article was written, it appears that the us has started favouring a ceasefire over continued hostilities.
The steady-state solution is unchanged
When we compute the steady-state solution, we find out it’s …
sum_p = sum(a.p() for a in model.actors) sum_px = sum(a.p() * a.x for a in model.actors) print(f'{sum_px/sum_p:.2f}')
0.35
… still 0.4. Seeing that we made significant changes to the set of actors we’re considering, that’s actually a promising sign. When we model the same dispute twice – using two different sets of actors – we get roughly the same answer.
The solution when two actors determine the outcome
However, the current outcome, the one we’ll see tomorrow and probably also the next week, is the minimum position of the Israeli government and Hamas. In other words, where one of those two actors set the bar for violence, that’s the level we’re stuck at. This is currently an unsurprising 0.2 – a continuation of the current level of hostilities.
This may be considered a variant of the existence of veto powers, i.e. actors that must agree for there to be an outcome at all. The difference in this case is that they don’t have to agree, but the outcome is the lowest of their preferences.
What we need to know, then, as a first step, are the chances that both of those actors consider a ceasefire.5 Once the ceasefire is in place, the peace brokering force will switch to desiring a full peace agreement, driving the situation along toward peace.
The attitude chart and its octants
In this type of modeling, Bueno de Mesquita often presents one actor’s outlook through a scatter plot. We will see precise details on the computations in a separate article, but for now, we’ll just say that each pair of countries has a specific 2D coordinate that represents their relationship in the dispute.
In line with what Bueno de Mesquita often does himself, I’ve reduced this 2D space into octants. If we look at the world from our point of view, we can see other actors A, B, C, D, E, F, G in these positions:
From our perspective:
C | B
|
D 2 | 1 A
-----------------AMBIVALENT-----------------
3 | 4 G
|
E | F
- Actors A and B are in quadrant 1. This quadrant is called the fight or crisis quadrant. A and B are the actors with which we may end up in a more serious conflict. (The difference between A and B is that we have the upper hand in a conflict against A, while B has the upper hand against us.)
- Actors C and D (those in quadrant 2) have the upper hand in an interaction
with us, but to different degrees:
- Actors C may try to get us to compromise and partially satisfy C’s demands, while
- Actors D may try to convince us to adopt their preference entirely.
- Actors E (those in quadrant 3) are those for which neither we nor they have any gain in interacting with.
- Actors F and G (those in quadrant 4) are those we have the upper hand aganist,
but to different degeres:
- We may try to convince actors F to adopt our preference entirely, whereas
- We may try to get actors G to compromise and partially satisfy our demands..
- Actors ambivalent are the ones which roughly agree with us so there’s no reason for us to interact.
One thing this rough version of the diagram does not capture is how intensely each actor desires to attain or avoid the interaction represented by a coordinate. This is useful information but I haven’t figured out a good way to include it in a lo-fi diagram like this. In lieu of a good way, I’ve chosen a bad way.
Each actor has four digits appended to the end of their name in the scatterplot,
with the first two digits being the utility of the challenging actor and the
last two being the utility of the actor from whose perspective we are looking.
The signs of the utilities must be inferred from the quadrant the point is in.
So for example, if a point is labeled IR1203, that might represent the Iranian
government has a +1.2 utility and the actor whose perspective we take has a
utility of -0.3.6 Can you tell I have been playing with openScope recently?
As a space-saving shortcut, if both numbers start with 0, the zeroes can be
omitted. So instead of writing PA0103 we may write PA13.
The current situation in Gaza
Using this, we can study the outlook of some key actors in the dispute at hand. Why not start with the Israeli government?
plot_situations(model, get_actor(model, 'IG'))
From the perspective of IG:
| IR00
|
|
-----------------------------Ih HA-----------------------------
| Id57 PA47 US00 PB01
|
|
This is read as such:
- The Israeli government may be preparing for a serious conflict with Iran, although neither party has much to gain from it in terms of this dispute.
- The Israeli government hopes to reach a compromise with the doves within its population to continue fighting at least a little bit.
- It has the same hope in regards to appeasing the Palestinian Authority, and they expect to barely gain anything from such a compromise with the us and the peace brokering force.
We’ll get back to this actor, but first, one that contains a surprising result:
plot_situations(model, get_actor(model, 'US'))
From the perspective of US:
IG00 IR40 |
|
|
------------------------------PB-------------------------------
| Id14 Ih01 HA01 PA14
|
|
Look at where Iran sits – in the compromise octant! This implies that if Iran challenges the us, then the us will not respond in kind but rather try some sort of compromise. Had I looked at this result earlier in the year I would not have believed it – but yet – on January 28 ago Iranian proxies struck us forces, and the us responded by handing Iran and their proxies such a long notice of their retaliation that they ended up “really giving it to those empty warehouses”, as one Twitter commentator phrased it. That’s a compromise if there ever was one.
That said, the us is very close to having positive utility against Iran in this conflict, so it may not require much to tip them over into conflict. Hm.
However, there’s one more interesting point on there: it seems like the us is powerful enough that if they challenged Hamas head on, they could get some sort of compromise out of them. On the other hand, if we look at the absolute value of the utilities involved, we see that the us has rather little motivation to do so.
Moving on, the Israeli doves are in a bad spot.
plot_situations(model, get_actor(model, 'Id'))
From the perspective of Id:
IG75 Ih52 HA54 US41 IR1902 PB41 |
|
|
------------------------------PA-------------------------------
|
|
|
They have the upper hand over nobody, are on fair terms only with the pa, and everyone else thinks they can wring a compromise out of the doves. In particular, Iran seems to think they have a lot to gain (+1.9 utility!) from angering the doves.
Evaluating changes to the situation
We can now perform what Bueno de Mesquita calls comparative statics, which is basically playing out various what if scenarios based on what we have seen so far. Here are some ideas:
- What if that conflict between Iran and Israel does happen and resolves in Iran’s favour, bringing Israel’s preference to Iran’s? What if it resolves in Israel’s favour?
- What if Iran gets a compromise out of the us?
- What if the US gets a compromise out of Hamas? (Unfortunately, this doesn’t change much because aside from determining the outcome, Hamas is not an influential force on other actors. This shouldn’t be surprising – the expected utility was low, after all.)
- What if Iran manages to anger the Israeli doves? (Not a whole lot, it turns out. The doves are just an easy target thanks to their low influence and edgy preference of full peace.)
Israel–Iran conflict resolutions
An interaction in the conflict/crisis quadrant has a highly uncertain outcome, and it can go either way. We’ll look at two of the extreme outcomes possible.
Israel adopts Iran’s preference
Since we have assumed Iran prefers intensified violence, and the level of violence is the highest desired by either the Israeli government or Hamas, Israel adopting Iran’s preference would result in intensified violence.
I want to reiterate that this doesn’t mean representatives of Israel sits down and calmly discusses an intensification of violence with Iran. A plausible way this could happen is through Iran launching attacks through proxies in Gaza, forcing the idf to retaliate in kind. The end result is the same: the idf steps us their violence.
We’ll quickly look at how the us perspective on the dispute changes from this.
explore_scenario(
model,
# Which actors get which new positions?
{ 'IG': get_actor(model, 'IR').x },
# From which actors' perspectives will we look at the result?
[ 'US' ]
)
0.0 ########## IG IR
0.1
0.2 ## Ih HA
0.3
0.4
0.5 ######## US PB
0.6
0.7
0.8
0.9
1.0 ### Id PA
From the perspective of US:
IG00 IR50 |
|
|
------------------------------PB-------------------------------
| Id14 Ih01 HA01 PA14
|
|
Not much happens except it becomes an even more profitable prospect for Iran to try to get the us to compromise off of their ceasefire position Since this is immaterial to the question we’re trying to answer7 What are the ways of achieving a ceasefire, rather than avoiding it. we will not look further into it.
Iran adopts Israel’s preference
If the conflict goes the other way, Iran would be forced to back down its support of proxies and accept a maintenance of the current level of hostilities.
This causes some small but interesting changes to relations between actors.
explore_scenario(
model,
# Which actors gets which new positions?
{ 'IR': get_actor(model, 'IG').x },
# From which actors' perspectives will we look at the result?
[ 'IG', 'Id' ]
)
0.0
0.1
0.2 ########## IG Ih HA IR
0.3
0.4
0.5 ###### US PB
0.6
0.7
0.8
0.9
1.0 ## Id PA
From the perspective of IG:
|
|
| US00
---------------------------Ih HA IR----------------------------
| Id57 PA56 PB01
|
|
From the perspective of Id:
IG75 Ih43 US41 IR1405 PB41 |
|
HA45 |
------------------------------PA-------------------------------
|
|
|
With this level of support for the current level of hostilities, Israel might become able to challenge the us to support them further than the us is currently interested in, and this would be a very contentious argument. It could result in the us forcing Israel to back the ceasefire position, which is a possible route to a ceasefire. Let’s remember that!
Now, here’s something I’m struggling to explain to the point where I’m inclined to say the model is wrong8 To be clear, I get exactly why the model mathematically arrives at this solution, what I’m trying to explain is what interpretation it could have in the real world.: when Iran takes a more passive stance, the model suggests that the Israeli doves would be more amenable to coercion from Hamas to accept the current level of violence. This is basically the model saying that de-escalation in the area results in doves being more belligerent. It certainly beats my intuition.
But it’s also not entirely nuts. Maybe, if Hamas commits another atrocity against Israel, the Israeli doves are willing to accept the current level of hostilities in Gaza – but only as long as they have some sort of reassurance it won’t escalate further. With Iran away from the extreme end of the scale, there are suddenly no powers in favour of escalation, and that might be the thing that allows the doves to accept the current level of hostilities.
I don’t know. Maybe the model is wrong, but it even if it is, it has helped us consider the dynamics of the situation in ways we otherwise may not have if left to our own devices.
Iran gets a compromise out of the US
This might sound good for Iran, but it’s not clear that it actually is. The second-order effects are that
explore_scenario(
model,
# Which actors gets which new positions?
{ 'US': 0.3 },
# From which actors' perspectives will we look at the result?
[ 'IG', 'US' ]
)
0.0 ######### IR
0.1
0.2 ########## IG Ih HA
0.3 ####### US
0.4
0.5 ##### PB
0.6
0.7
0.8
0.9
1.0 ##### Id PA
From the perspective of IG:
|
|
| US00 IR00
-----------------------------Ih HA-----------------------------
| Id57 PA57 PB11
|
|
From the perspective of US:
| IG00 IR20
|
|
---------------------------------------------------------------
| Id16 Ih00 HA00 PA15 PB00
|
|
In other words
- Israel gets a slight upper hand in future conflicts against Iran (because of stronger support from the us in favour of military interventions), and
- The us will then no longer be satisfied with compromising to Iran but will instead ready themselves to further escalate the dispute.
The latter goes against my intuition: my feeling suggested the us would leave the dispute at the current retaliation effort, but the model suggests it may increase. But if I think about it, this outcome represents the us getting dragged further into the conflict, which surely increasing their likelihood of engaging with all other actors.
The peace brokers get a compromise out of Israeli doves
This does not change all that much, because the Israeli doves are not very influential according to these assumptions. However, a compromise would make them more vulnerable to challenges from several other strong actors.
explore_scenario(
model,
# Which actors gets which new positions?
{ 'Id': 0.8 },
# From which actors' perspectives will we look at the result?
[ 'Id' ]
)
0.0 ######## IR
0.1
0.2 ######## IG Ih HA
0.3
0.4
0.5 ########## US PB
0.6
0.7
0.8 ## Id
0.9
1.0 ## PA
From the perspective of Id:
Ih33 IR1102 |
|
IG55 HA24 US12 PB12 |
---------------------------------------------------------------
| PA14
|
|
For example, once they have compromised with the peace brokers, the peace brokers will think they can get them to fully support a ceasefire, rather than the full peace they would have wanted from the start.
Iran backs down, Israel yields to US
Remember how, when Iran stopped escalating the situation and accepted the status quo, the us might have a shot at convincing Israel that a ceasefire is appropriate? What happens if they succeed?
explore_scenario(
model,
# Which actors gets which new positions?
{ 'IR': 0.2, 'IG': 0.5 },
# From which actors' perspectives will we look at the result?
[ 'IG', 'HA' ]
)
0.0
0.1
0.2 ###### Ih HA IR
0.3
0.4
0.5 ########## IG US PB
0.6
0.7
0.8
0.9
1.0 ## Id PA
From the perspective of IG:
|
|
|
-----------------------------US PB-----------------------------
| Ih12 HA12
|
| Id54 PA54 IR00
From the perspective of HA:
IG21 US11 PB11 |
|
|
-----------------------------Ih IR-----------------------------
|
|
| Id43 PA43
At that point, the Israeli governments negotiating position with its doves and the pa becomes very strong. This lack of support for continued violence is enough to try to get Hamas to compromise closer to a ceasefire as well.
explore_scenario(
model,
# Which actors gets which new positions?
{ 'IR': 0.2, 'IG': 0.5, 'PA': 0.5, 'Id': 0.5, 'HA': 0.3 },
# From which actors' perspectives will we look at the result?
[ 'HA' ]
)
0.0
0.1
0.2 ##### Ih IR
0.3 HA
0.4
0.5 ########## IG Id PA US PB
0.6
0.7
0.8
0.9
1.0
From the perspective of HA:
IG11 Id40 PA40 US10 IR30 PB00 |
|
|
---------------------------------------------------------------
| Ih00
|
|
With this level of support for a ceasefire, most actors have a chance of issuing credible enough thretas that Hamas compromise even further toward a ceasefire.
What happens after a ceasefire
Although we haven’t found an easy way to a ceasefire with the current set of actors snd priorities, we can pretend for a second that we have, and try to explore what the next steps would be.
In other words, both the Israeli government and Hamas now support a ceasefire, and then the peace brokering force and the us switch to supporting a full peace agreement.9 Iran has also left their extreme position, because I don’t see a ceasefire as possible unless that happens.
explore_scenario(
model,
# Which actors gets which new positions?
{ 'IR': 0.2, 'IG': 0.5, 'HA': 0.5, 'PB': 1.0, 'US': 1.0 },
# From which actors' perspectives will we look at the result?
[ 'IG', 'PB' ]
)
0.0
0.1
0.2 ####### Ih IR
0.3
0.4
0.5 ##### IG HA
0.6
0.7
0.8
0.9
1.0 ########## Id PA US PB
From the perspective of IG:
| Id22 PA32 IR10
|
|
------------------------------HA-------------------------------
| Ih12
|
| US21 PB22
From the perspective of PB:
IR63 |
|
IG22 HA11 |
---------------------------Id PA US----------------------------
|
|
Ih31 |
Oddly enough, the peace brokering force is still rather weak. They consider themselves vulnerable to a challenge from all three main belligerent actors (Israeli government, Hamas, and Iran) and is not obviously able to hold firm in its commitment to full peace. The Israeli government thinks of itself as capable of convincing both the us and the peace brokering force that a temporary ceasefire is what’s best for the region.
Conclusion
This is the part that still takes some subjective judgment because we have not yet achieved the full dynamic actor–motivation model. Looking back at the comparative statics we played, we suspect that:
- If the us gets more involved – which can happen due to Iranian attacks – Israel becomes a more dominating force in the region and the dispute will escalate.
- Iran disengaging slightly because of combat losses would increase the support of a caesefire. (Both by making it easier to convince the hawks, and – contrary to intuition – by making the doves more accepting of the current level of violence.)
- The peace brokering force is weak however we look at it, and there is no clear way toward a ceasefire with this set of actors. If it happens, though, it would be because the us exploits the moment where Iran has stepped back slightly, and manages to wrest a concession from an Israeli government that it wasn’t interested in until it was forced to. But this conflict would be contentious and the outcome would be unusually uncertain.
- Then even if we did get to a ceasefire, peace is a distant prospect after that.
So all in all, a peace agreement in the near term seems very unlikely. Weighing together the above, I would judge the following conditional probabilities to be necessary, considering that it all has to happen within the next four months:
- Iran disengages slightly: 50 %.
- The us exploits that moment and overpowers the desires of the Israeli government: 40 %.
- Ceasefire turns into a peace deal: 20 %.10 This seems even less likely than that, but there’s a lot of ludic uncertainty involved so it would be unwise to be too confident.
In total, that’s 4 %. We would be wise to adjust upward because four months is a long time and the actors may change during that time, so let’s say 12 %.
This is much more aggressive than our previous gut feeling of 49 %, but then we have to remember that forecast was issued about two months earlier when uncertainties were indeed larger. Also, I was probably optimistic even then – I think I have a general peace bias.