| Chapter 27 | |
| Tools for Reasonocracy | |
| We assert that reasoning is a better process for decision making that power. LetÕs put some stakes in the ground to get practical. | |
| Goals | |
| Our highest-level goal for government is the greatest good for the greatest number with minority rights. How might we compute that? | |
| Good-for-each-individual x number-of-individuals | |
| This is a simple metric, but often too simple. Suppose we have 10 individuals. Policy A changes the income of our richest individual from $100K per year to $200K per year but decreases the income of the other 9 from $50K per year to $45K per year. Is it a good policy? Total wealth gained is $100KÑ(9 * 5K) = $55K clearly a net positive. | |
| But we are comparing different wealth individuals, so another way to analyze it is to add up the percentage gains: Our rich guy goes up in his wealth by 100% (doubling his income), whereas each of our 9 go down by 10%. 100-percent-pointsÑ90-percent-points = 10-percent-points, still a net gain but not as much of one. A 3rd way to analyze this is that 1 guy makes more and 9 make less, so if we balance our winner with 1 loser, our net is 8 losers. So counting Òwhole team scoreÓ, policy A is a big win, but when counting individuals, its a big loss. | |
| What does Capitalism do? Well neither really. Capitalism tends to make it easier for the rich to get richer and the rich make decisions (like tax rules) to benefit themselves. So Capitalism keeps score in a 4th way: If the rich (few) do well, thatÕs good; everybody else doesnÕt matter much. Note that our 4th way happens to correspond to an overall higher Òteam scoreÓ in the above example but thatÕs an accident in Capitalism, not a designed-for criteria. Had the poor people lost more than the rich guy gained, team score would be negative but Capitalism would still be for it. | |
| Our example is not so hypothetical. Especially in the last decade, US GDP is up, wealth of the wealthiest 1% is up, but poor peopleÕs income is down. With 2 benefits, the status quo can claim its maximizing team wealth and back that up with overall GDP numbers, but the correlation between Policy A and team wealth is not causally linked. | |
| Bhutan [Bhutan 2016] measures success by ÒGross National HappinessÓ. This is more complex than GDP, so its success is more complex to assess. But its a pretty compelling idea, with the caveat of how to manage the complexity. | |
| The DevilÕs in the details | |
| Gross measurements of dollars or happiness are crude tools. And assessment is only part of the story, as any opponent of ÒNo Child Left BehindÓÕs educational testing policy can tell you. Measurement alone does not provide new solutions. | |
| Negotiation | |
| Some of our most powerful tools are conceptual. Such rather loose concepts can be packaged into processes that make them actionable. In Good for You, Great for Me, Larry Suskind [Susskind 2014], one of our favorite MIT professors. Teaches how to make negotiations more beneficial for both (conßicting) parties. An example is in order. | |
| Many cities have laws limiting the number of liquor licenses they will grant. A restaurant must have one to sell alcohol. Our example city has distributed all the allotted licenses, yet 2 unfriendly business men each want one. A restaurant owner retires and the license reverts back to the city. The two business men each want the license. | |
| Scenario 1: Each presents to the city council why they should be given the license. Joe hires a pro marketing firm that designs presentations and wins. | |
| Scenario 2: Sam learns that Joe has hired a marketing team so he does the same thing. Then Joe hears about this and gives his team more money. Joe and Sam are now in a classic Òarms raceÓ where the real winners will be the arms merchants (in this case marketing firms). Eventually either Joe or Sam will win but the victory will be reduced by the marketing expenses. | |
| Scenario 3: Joe learns Sam is also employing marketers and realizes they are in an arms race so he fires his marketing staff and uses his marketing money to bribe the councilmen. By switching ÒmediaÓ he hopes to avoid an arms race, but just creates another as Sam switches to the same tactic. | |
| Scenario 4: Sam realizes it doesnÕt matter which battlefield the arms race is on (marketing or bribery), it still drains profits of both combatants. He convinces Joe to pay the other half of the fee of a Negotiation facilitator named Larry. LarryÕs strategy is: | |
| 1. Think of some new options that neither Joe or Sam could come up with on their own. | |
| 2. Help them collectively choose the one with the best outcome, trying for a win-win in a situation that originally seemed like it had to be a win-lose. | |
| Scenario 5: The first new card Larry puts on the table is maybe we can convince the city council to split the original 1 full liquor license into two beer-and-wine-only licenses. Estimating from other businesses, they determine each would get about 40% of the revenue of the full liquor license. | |
| Scenario 6: Larry doesnÕt like losing 20% of the revenue. Joe and Sam arenÕt happy about it either, yet it still sounds more attractive than a 50% chance of getting nothing. Next solution is to have Joe and Sam co-own one new restaurant, sharing costs and profits. Now each is getting 50% of the profit so Òthe teamÓ isnÕt losing 20%. This is even better. Each spends 50% of what they would have to get the restaurant by themselves, and each gets 50% of the profit. | |
| Scenario 6: In trying to seal the deal, Larry realizes that each is a little hesitant: they each are bothered by not getting that 100% of the profits. But they each still have half the capital to start a 2nd restaurant. So Larry proposes they form a ÒchainÓ and explore liquor licenses in nearby towns to also co-invest in. Each investor likes the increased diversity this new approach affords, making the total investment more stable. Deal Sealed. | |
| What can we learn from this? | |
| 1. By getting creative, part of the wealth of a winner can compensate a loser. This technique is especially important in the Òcut the baby in halfÓ compromise situations. | |
| 2. By cooperating, an expensive battle can be avoided, benefiting both parties. | |
| 3. By Òexpanding the pieÓ a Òdivide down the middleÓ compromise doesnÕt have to mean that each party gets a smaller piece. | |
| Negotiation is generally restricted to a few parties. But a government has millions of parties to appease and a myriad of diverse issues to resolve. We need some new tools. | |
| Justify | |
| To help people understand how their own ideas relate to the ideas of others, we wrote a program called Justify. Justify is a language for expressing rationale, and a Òdevelopment environmentÓ for organizing those ideas to help people model complex situations and make better decisions. Justify tries to encapsulate each idea or opinion in what it calls a point. Roughly, a point is what you might write in a single post in an online discussion. Justify guides people to declaring what kind of ÒpointÓ they are making (one of a couple hundred different kinds) and where that point should go in a hierarchy of related points. | |
| This architecture doesnÕt restrict what can be said, nor when something can be said, but by restricting where and how something is said within a hierarchy, it does help mold complex ideas into a coherent discussion. Justify also summarizes each point and its sub-points with an automatically created assessment. This helps humans understand the implications of their reasoning at each level, and helps the program make even higher-level assessments. Navigating a complex hierarchy is facilitated by the user seeing, at each level, an assessment of lower-level points [Fry 2013]. | |
| We propose Justify as a standard venue for governmentÕs decision making process. JustifyÕs assessments donÕt depend on who said something (like power-based governments), but rather on what was said and especially why it was said. Because points are organized semantically, not chronologically, the order that statements are made in is unimportant, unlike the debates and ordinary discussions of democracy and consensus. | |
| Above we outlined how a skilled negotiator can find opportunities for value for two conßicting parties and reach agreement. Sadly, skilled negotiators like Larry Suskind are rare. Can we make a tool that might lead participants to a similarly positive outcome? | |
| Language | |
| We can steal a winning idea from programming: To solve a problem, first define a language to describe problems, then use that language to describe particular problems and their solutions. Programmers use general purpose programming languages, like Python, and JavaScript, to model the underlying issues of their applications and to create new facilities to work with those issues. But programming is complex, so special programs called Òdevelopment environmentsÓ have been created to facilitate working with programming languages to create new programs. | |
| Part of Justify is a new language to describe rationale for making decisions. Like a general purpose programming language, it is not targeted at any particular kind of decision. It tries hard to be Ògeneral purposeÓ enough for any kind of decision. A small group discussion can use JustifyÕs point type language to express the role that each of their contributions plays in the overall argument. Justify is implemented in the general purpose programming languages of Python and JavaScript, and runs on the web, so anyone can use it. http://justify-app.appspot.com/. | |
| A tale of two governments | |
| LetÕs walk though a concrete example with two different processes. Our example is in deciding the material for a new bridge. The bridge will provide transport for a million cars a year and hopefully last decades. Millions of dollars are at stake. The transportation committee has narrowed down our viable options to two: carbon fiber and (we said it was a concrete example) concrete. LetÕs listen in. | |
| Power-based government scenario | |
| In the power-based discussion, we hear lots of reasonable sounding rationale. It comes from the companies pushing product, the legislators and the presumably neutral university experts. This material has a proven track record. That material is stronger. This oneÕs cheaper, that one is corrosion resistant. This new formula overcomes the previous defects, and so on. Some of it is accurate, some of it isnÕt. | |
| But it really doesnÕt matter because, in conventional politics, itÕs not a reason-based decision making process. Behind the scenes dollars are passed, promises made and deals are sealed. In a complex situation, you can rationalize anything. The side with the most under the table wins. Yes they need to have the pedigreed experts, the ÒconservativeÓ spreadsheet, the Òmakes senseÓ sound bite. But the other side has all that too. | |
| What it comes down to is the most cash (or the equivalent) in the right pocket at the right time. Power based governments back the status quo. Concrete wins over carbon fiber. Innovation occasionally wins, but only occasionally. | |
| Reason-based government scenario | |
| Superficially, a reason-based process can look very similar. Expert testimony, rational, rebuttals, etc. The difference is that what is said actually matters. There isnÕt a back room where the deal is sealed. | |
| Because a reason-based process would allow for learning, itÕs likely that we would come up with possibilities beyond the original two. Our final design gets the benefits of carbon fiber where its needed most, and the benefits of concrete where that is most appropriate. We donÕt end up with any big winners. Manufacturers are compensated fairly with a profit that allows them to do their best work, but not retire. Citizens get a fair deal for their tax dollars, not something that sounds too good to be true. Legislators had to work hard to hammer out the deal, but didnÕt have to break laws, wallets or hearts to do it. | |
| With these themes in mind, youÕre prepared for the next level of detail. We frame our issue with a question. | |
| Our Justify point has a title representing the question. The pink Ò?Ó represents the point type. The ÒAssessmentÓ is Òunfinished because there is no rationale for answering the question, indicating that we need more detail. Our first rational is one in favor of concrete: | |
| We have added a ÒproÓ point that assesses to itself and, because its the only reationale we have, becomes the assessment of our question indicating that concrete is best. However, the carbon avocates rebut: | |
| Our new con point assesses to itself, and causes the assessment of ÒYes its cheaper per pound.Ó to become ÒrefutedÓ. Because that 2nd point is refuted, it no longer counts for our top level assessment, which reverts to ÒunfinishedÓ. An open minded participant asks an honest question. | |
| Because this question has no possible answers, it is Òunfinished and that ÒunfinishedÓ propagates all the way to the top. Letting us know we have work to do. | |
| Giving our 2nd question an answer makes it no longer unfinished, and restores the refutation of ÒYes its cheaperÓ 3 levels up. WE have more information but still no resolution to our top question. | |
| Our last point is not addressing the ÒcheaperÓ aspect, but going for a new rationale immediately under our top level question. Now our top level question has one con point beneath it and no other unrefuted points, so its assessment is ÒconÓ, i.e. concrete is not better than carbon fiber. | |
| Complexity | |
| In a 6 point discussion, humans can keep track of the rationale and what it Òadd upÓ to. But imagine the hundreds of points in a real debate about safety, durability, capacity all with their own cost and aesthetic considerations. When humans are faced with overwhelming complexity, they tend to throw out rationale and Ògo with their gutÓ. In a complex world, such emotional strategies cause sub-optimal choices. | |
| Justify comes up, not simply with the top level answer, but answers to embedded questions and the implications of rationale at each level. Its reasoning is transparent for all to see. If a new situation arises, say a new high strength concrete is developed, we can revisit our discussion without having to start all over. The information about carbon and aesthetics is still valid. Other points can be placed in the new context, not by changing those points, but by adding new rationale. Thus our historical reasoning is preserved, yet refined to a more accurate picture of the new reality. This allows new entrants to our discussion to both understand what preceded them, and add to it in a way that builds greater understanding. | |
| Justify enables a user to not have to wade through volumes of arguments that they donÕt care about, yet facilitates diving down into the details of those points they do care about. Those triangles to the left of each point allows shrinking or expanding the points under any point. This permits not just a superficial high level view, or an Òall detailsÓ view, but rather, detail where you want it and not where you donÕt, all a few mouse clicks away. | |
| Can we afford government by reason? | |
| Above weÕve presented two different techniques that can improve upon Òemote and voteÓ. Each requires more effort than our typically simplistic voting schemes. (Or does it? The US 2016 presidential election cost billions of dollars.) A more reasoned approach, even if it made government just a few percent better, would surely pay for itself. | |
| An innovation mind-set would encourage inventing new techniques for improving the reasoning capacity of decision makers. The development and deployment of these advances needs education, the will to improve our world, and persistence in the face of a repressive status quo. Expensive to be sure. But, we believe, cheaper than the alternatives. |