Wicked Problem Attempt: World Peace

Avoiding another world war

(Request access to code through email)

Evaluation Function

For our evaluation function we used the probability of nuclear war equation from the Global Catastrophic Risk institute1. We felt that this equation would be a good model for our problem because we live in a world where most of the world’s powerful nations have access to nuclear weapons and if another world war were to happen, it would likely be nuclear. The probability for any war occuring is very hard to quantify, so we utilized a simplified version of the equation they defined. The Global Catastrophic Risk Institute came up with the probability using a list of historical events and nuclear war scenarios such as the Cuban missile crisis during the cold war. With λT being defined as the annualized rate of events that could provoke intentional first strike (foreign policy issues). While λF defines the annualized rate of a false belief of being under attack (domestic issues). Each being multiplied by the conditional probability that one such event would lead to first strike. The events and beliefs that the GDR institute used were all the past to current historical incidents between nuclear countries. Getting such data would not help model the future reduction of another world war occurring. We decided to simplify these into scores of different states of the world that can represent our two λT and λF equations.

 def q(self):
        # Composite Scores
        self.gammaT = i + f
        self.gammaF = e + h

        # Qscore
        self.q = pt(self.gammaT) + pf(self.gammaF)
        return self.q

pt = lambda x: x / 100
pf = lambda x: x / 100

Model: Q(s): λnw = λT * PT + λF * PF

State Information

Scores:
	I - Ideological score
	
	F - Foreign Policy Score
	
	E - Economic Score, includes budget increases back up yearly
	
	H - Hysteria Score

λT : is the annualized rate of events that could provoke intentional first strike. Examples of such events include:

Religious disputes (Crusades, ISIS, etc.)

Ideological Radicalism (Communism, Nazism)

Economic Disputes (Trade Wars, Sanctions)

Preventative Wars (Shut down totalitarian states, First strike advantage)

We model λT as I + F

λF : is the annualized rate of a false belief of being under attack. Issues that could impact this include:

Economic Issues (Inflation, Poverty, National Debt)

Lack of Democracies (Democracies are less likely to go to war with each other)

Hysteria (Social Media, fake news)

We model λF as E + H

Example States

Funds: 500 Income: 5 Basis Scores: ideo score: 5 fp score: 5 econ score: 5 hysteria score:
 5  GammaT: 10 GammaF: 10 Qscore: 0.2
 
Funds: 500 Income: 5 Basis Scores: ideo score: 3 fp score: 3 econ score: 3 hysteria score:
 3  GammaT: 6 GammaF: 6 Qscore: 0.12
 
Funds: 500 Income: 5 Basis Scores: ideo score: 10 fp score: 10 econ score: 10 hysteria sco
re: 10  GammaT: 20 GammaF: 20 Qscore: 0.4

Funds: 500 Income: 5 Basis Scores: ideo score: 8 fp score: 5 econ score: 1 hysteria score:
 10  GammaT: 13 GammaF: 11 Qscore: 0.24

Funds: 500 Income: 5 Basis Scores: ideo score: 1 fp score: 5 econ score: 10 hysteria score
: 1  GammaT: 6 GammaF: 11 Qscore: 0.1699

Operators

Lifting Tariffs

Boosts F (improves relations with the nation in question)

Hurts E (can be bad for domestic companies)

Backing a Coup (Harsh chance of increasing tensions between countries) 

Boost I (chance of toppling radical leader)

Hurts F (bad public image, can make other nations angry)

Hurts H (perceived as dangerous by the public)

Costs $

Invest in an education campaign

Boosts H (more educated people are less likely to become hysteric because of the news or be persuaded by fake news)

Boost I (more educated people are less likely to be susceptible to radicalism)

Cost $

Raise Taxes (Budget up, hurts economic score)

Hurts E (bad for people and businesses)

Increase $

Lower Taxes (Budget down, helps economic score)

Boosts E

Decrease $

Candidate Solution 1

The current state

Action taken to get to that state.

Funds: 500 Income: 5 Basis Scores: ideo score: 5 fp score: 5 econ score: 5 hysteria score: 5  GammaT: 10 GammaF: 10 Qscore: 0.2

Start State: The start state is based on the current state of the world and the likelihood of us heading to another world war. According to the global conflict tracker8, other than the middle east we are in an unprecedented time of peace. Since each of the scores, ideology, economic, hysteria, and foreign policy, can be in a range from 0 to 25 we guesstimated that 5 would be a good starting position. Giving us a wiggle room to lower our Q score, while also not being so high that it does not make sense.

Funds: 404.73103819257665 Income: 5 Basis Scores: ideo score: 0 fp score: 5.249038710577172 econ score: 5 hysteria score: 5.249038710577172  GammaT: 5.249038710577172 GammaF: 10.249038710
577171 Qscore: 0.15498077421154344

backCoup North Korea:

Backing a coup in North Korea decreased the ideology score to zero. We calculated the decrease in the ideology score by subtracting the current score by the inverse of the democratic score multiplied by the power of that nation. Since North Korea ranks high on the power index, while also ranking low on the democratic score if the coup were to be successful it would reduce the ideology score by a large amount. Reducing the ideological craze in the world. Backing a coup also increases the hysteria score and foreign policy score by a fraction of the modifier that was used to reduce ideology score. Backing coup has 20 different variations, since it can be applied to every country that scores below a 5 on the democratic index. We also limited the game so that a coup can only be done once for every country.

Funds: 359.60534689041475 Income: 5 Basis Scores: ideo score: 0 fp score: 3.363669178148816 econ score: 5.942684766214178 hysteria score: 5.249038710577172  GammaT: 3.363669178148816 GammaF: 11.19172347679135 Qscore: 0.14555392654940164

liftTariff Sweden

Lifting tariffs on Sweden decreased the foreign policy score based on the democratic index value of Sweden while increasing the economic score. Since the US has tariffs on most of the world’s countries, the liftTariff action can be enacted on every country on the list of countries once. Lifting tariffs is used as a political strategy to show good faith to countries following the players interests. Since foreign policy could be its own wicked problem, the current command does over simplify the action as it relates to enacting peace.

Funds: 359.60534689041475 Income: 21.82741116751269 Basis Scores: ideo score: 0 fp score: 3.363669178148816 econ score: 7.131221719457014 hysteria score: 5.249038710577172  GammaT: 3.3636
69178148816 GammaF: 12.380260430034186 Qscore: 0.15743929608183002

RaiseTaxes

Given that the modeled actions are based on what a country could do to prevent another world war, the yearly income can be increased by raising taxes. This would cause the economy to worsen for the player country, leading to diminishing returns when further increasing taxes. In this case raising taxes increased the income from 5 to 21 million per year.

Funds: 281.43275805792746 Income: 21.82741116751269 Basis Scores: ideo score: 0 fp score: 3.363669178148816 econ score: 7.131221719457014 hysteria score: 3.249038710577172  GammaT: 3.3636
69178148816 GammaF: 10.380260430034186 Qscore: 0.13743929608183003

Educate

The player can choose to educate the population in order to reduce the hysteria and ideology score. Since education provides no other down sides, the cost of enacting it was set to 100 million dollars. The hysteria and ideology score is lowered by 2 when this action is called, being capped to 0.

Funds: 281.43275805792746 Income: 7.804568527918782 Basis Scores: ideo score: 0 fp score: 3.363669178148816 econ score: 5.942684766214178 hysteria score: 3.249038710577172  GammaT: 3.3636
69178148816 GammaF: 9.19172347679135 Qscore: 0.12555392654940167

lowerTaxes

Lowering taxes as an action is a domestic policy in order to improve the economic score. It would result in a decrease in income for any following turns, for example from the previous state to this state the income went from 21.8 to 7.8. The new income is calculated by the given function (current income - (15 / econ score)). The function itself is rather arbitrary, since modeling the economy is in itself a wicked problem. We attempted to best capture how the state of the economy might affect the budget allocated to the policies that the players can make.

Candidate Solution 2

CURRENT_STATE = startState
Funds: 500 Income: 5 Basis Scores: ideo score: 5 fp score: 5 econ score: 5 hysteria score: 5  GammaT: 10 GammaF: 10 Qscore: 0.2

CURRENT_STATE = liftTariff China
Funds: 359.0352146647371 Income: 5 Basis Scores: ideo score: 3 fp score: 0 econ score: 12.235890014471781 hysteria score: 3  GammaT: 3 GammaF: 15.235890014471781 Qscore: 0.182358900144717

CURRENT_STATE = educate
Funds: 264.0352146647371 Income: 5 Basis Scores: ideo score: 1 fp score: 0 econ score: 12.235890014471781 hysteria score: 1  GammaT: 1 GammaF: 13.235890014471781 
Qscore: 0.14235890014471783

Summary

In our problem formulation we attempted to capture a diverse set of strategies for avoiding another world war by taking into account how the United States (our agent) might interact with different countries by levying tariffs for economic pressure and backing coups for military pressure. We also wanted to take into account possible domestic policies so we allowed the agent to invest in education campaigns and raise/lower taxes to control its funds. Our aim in doing this was to gain some insight into what kind of strategy would be best for the US in the real world.

In practice, the agent is able to interact with a large set of foreign countries and the impact of their interaction with a given country is dependent on the power and democracy score of that country. This way, interacting with Russia or Iran has a much more dramatic impact than interacting with Peru, for example. If a country is more powerful, it will be more expensive and more dangerous to back a coup there, but the potential reward of doing so is higher. As an example, it would probably be fairly risky and expensive to back a coup in North Korea, but the potential reward of toppling a dangerous dictator like Kim Jong Un could be fairly high. The same vein of logic applies to applying pressure in the form of tariffs - it is risky to apply too much pressure on a powerful nation like China, but if done properly, it could prevent them from becoming too powerful. When it comes to war, especially nuclear war, education is of high importance. Educated nations are less likely to declare war on eachother, become susceptible to radical ideologies, or become hysteric in reaction to the news. All of these factors can impact the likelihood of a world war so we wanted to allow the agent to invest in education as a domestic policy to moderate these things. Of course, all of these described policies have a financial cost so we also allow the agent to raise and lower taxes to control its access to funds. Our strategy for calculating the impacts that the policies would have was to keep track of four metrics: ideological threats, foreign policy threats, economic threats, and hysteria. We chose these metrics because they were considered major factors listed in the Stanford Reasons for Wars survey and we felt that they fit well with the equation we used from the Global Catastrophic Risk Institute’s Model for the Probability of Nuclear War.

We believe we were fairly successful at constructing the problem using the formulation described above but we ran into some issues when determining what numbers to use for the relative impact that the policies would have on the metrics. It was very difficult to find good information on how to quantify the impact of enacting the described policies and as a result, many of the numbers we used other than the power index and democracy scores are based on our own educated guesses. This is definitely the weakest aspect of our model and we believe it had a reasonable impact on the solutions found by the AI. For example, we noticed a tendency for the AI to focus primarily on domestic policies rather than foreign policies and believe that this may be influenced by the constants we chose rather than what solution is truly ideal. We also ran into trouble when defining more ambitious goal states because the computation took too long. We believe this is primarily because with all the different nations that the agent can interact with, there are almost 100 distinct operators and after a certain depth the algorithms get bogged down.

Overall, we feel that our problem formulation was a success although we acknowledge that there are many ways it could be improved.

Bibliography

“2020 Military Strength Ranking.” Global Firepower - World Military Strength, www.globalfirepower.com/countries-listing.asp.

Baum, Seth, et al. “A Model for the Probability of Nuclear War.” SSRN Electronic Journal, Global Catastrophic Risk Institute Working Paper 18-1, 2018, doi:10.2139/ssrn.3137081. https://www.ssrn.com/index.cfm/en/

“Democracy Index.” Wikipedia, Wikimedia Foundation, en.wikipedia.org/wiki/Democracy_Index.

“Federal Spending: Where Does the Money Go.” National Priorities Project, www.nationalpriorities.org/budget-basics/federal-budget-101/spending/.

Jackson, Matthew O., and Massimo Morelli. “The Reasons for Wars: An Updated Survey.” The Handbook on the Political Economy of War, Dec. 2009, doi:10.4337/9781849808323.00009. https://web.stanford.edu/~jacksonm/war-overview.pdf

Lynn-Jones, Sean M. “Why the United States Should Spread Democracy.” Belfer Center for Science and International Affairs, Mar. 1998, www.belfercenter.org/publication/why-united-states-should-spread-democracy.

“Sanctions Programs and Country Information.” U.S. Department of the Treasury, www.treasury.gov/resource-center/sanctions/programs/pages/programs.aspx.

“Global Conflict Tracker l Council on Foreign Relations.” Council on Foreign Relations, Council on Foreign Relations, 22 Apr. 2020, www.cfr.org/interactive/global-conflict-tracker?category=us.