Three Kinds of Compromise
The other day I was writing something about how Approval Voting, like other cardinal (rating) voting systems, produces compromise results, meaning results more likely to incorporate the opinion of all members of a group—even minorities. To spice things up I searched the web for a pithy compromise quote, and was appalled to find that most were negative. It turns out that this is not a new phenomenon, as Theodore Roosevelt observed in 1900:
“‘Compromise’ is so often used in a bad sense that it is difficult to remember that properly it merely describes the process of reaching an agreement. Naturally there are certain subjects on which no man can compromise. For instance, there must be no compromise under any circumstances with official corruption, and of course no man should hesitate to say as much.”
A web search on compromise+dirty+word will give you the picture.
The original, now archaic meaning of compromise was for two disputing parties to commit (co-promise) to submit to arbitration, the result of which could be expected to fall somewhere between their negotiating positions. The primary modern meaning is now to make concessions or tradeoffs to find such a mutually acceptable middle solution. Such compromise is positive, unlike the secondary modern meanings: to make excessive concessions out of expediency or dishonesty (e.g. "compromise one’s principles"), or to damage something (e.g. one’s reputation or a security system). Under its primary meaning, compromise is essential to human interaction, peace, and prosperity, so it is quite sad to see the word tainted by negative secondary meanings. There is no adequate synonym ("consensus" has been suggested to me, but that is something else). Perhaps if more people had a deeper understanding of compromise, we could rid the word of its negative connotations.
First let us get the negative meanings out of the way. Engineers are well acquainted with design compromises—we call them solutions! Many requirements—such as those for features and cost—are incompatible, so the designers search for a solution that optimally balances the requirements with minimal trade-offs. If someday in the future that same system gets damaged or hacked, some might call it “compromised”, but instead of this stilted usage we could just as easily say damaged or hacked.
Of course it is not good for something to fail to meet its minimum requirements, whether by original intent or later damage—thus the negative connotations of compromising on fundamental principles. But a declaration like “We never compromise on safety!” is mere talk—safety has a cost, and is just as subject to compromise, albeit with a high weighting. Defiant statements about never compromising (I found several compromise quotes by Margaret Thatcher and Ayn Rand, and in 2011 NPR had a story about House Speaker John Boehner’s inability to even speak the word compromise) are just bravado to rally the troops or impress the enemy. We are talking about maximizing utility within constraints. We should call over-compromising overcompromising, especially if one is negotiating on behalf of someone else and one's concessions are motivated by weakness, uncertainty, incompetence, or personal gain (e.g. when a politician is caught in a compromising position). What are the proper motivations for compromise?
Group decision-making considers three fundamental qualities—interest, knowledge, and power—when combining the group members’ individual opinions. This triad shows up in other areas, e.g. legal theory (see Randy Barnett’s The Structure of Liberty), and it is not surprising that motivations for compromise fall into the same categories. Compromise occurs when we recognize and accept the differing interest, knowledge, and/or power of fellow decision-makers, and we make appropriate concessions. The reasons for our acceptance could be a higher-level shared interest, humility about the limits of our own knowledge, or realism about the limits of the possible.
To the degree that voters (or decision-makers) have diverging interests, a group decision is less likely to make them all happy. The group decision-making method called Analytic Hierarchy Process contains an explicit step where the participants rank their evaluation criteria, producing a measure of the group's interest alignment. If their interests are highly divergent (e.g. the group consists of wolves and sheep), they should question why they are together, and whether a single group decision is realistically possible.
However, most people who find themselves in a group and making a group decision probably do share some interests. A group of friends deciding on a restaurant or movie has an overarching interest in remaining friends, so it is unlikely that a majority of them would impose an unpleasant decision on a minority. They would naturally use a decision-making system that incorporates the opinions of all group members, and produces compromise in one of its four manifestations: the proportion of a multi-winner decision, the qualities of a single-winner decision, rotation (time-sharing), and horse-trading (linking two or more decisions).
Voters, like participants in a market, possess imperfect information, principles, and judgment. To the degree that they are trying to reach a “true” decision (e.g. estimating the number of jellybeans in a jar), they should value the information contributions of other voters, and acknowledge their own limitations. This is an example of epistemic humility, one of the epistemic virtues. Harder for most people to admit is that their judgment and principles, e.g. their values, are also flawed.
Recent research has shown that numerous cognitive biases reliably lead human beings to commit errors in reasoning. Worse, our reasoning appears to be employed not independently but selectively in support of our interests (so-called motivated reasoning), which we perceive intuitively. Daniel Kahneman, the psychologist best known for cataloguing cognitive biases, stated that the worst of them—the deadliest epistemic sin—is overconfidence.
Even our most fundamental values are suspect. Moral Foundations Theory has shown that human beings evaluate moral problems according to six moral foundations, analogous to the five basic tastes on the tongue. Same as people are born with different taste preferences (e.g. conservative or novelty-seeking), they are born with different moral-foundation sensitivities.
Research (described in Jonathan Haidt’s book The Righteous Mind) shows that the three main political camps (conservatives, progressives, and libertarians) cluster measurably in different regions of moral-foundation space: conservatives on average are equally sensitive to all six moral foundations, progressives sensitive primarily to Care and Fairness, and libertarians sensitive to Liberty and Fairness. Switching metaphors, it is as though conservatives see normally, while progressives are blue-yellow color blind and libertarians are red-green color blind.
Each type of perception has its advantages and disadvantages, and so persists in the population in stable proportions. The theory suggests that the political camps originate in the ancestral human environment, with conservatives focused on intertribal competition, progressives on intra-tribal competition, and libertarians on excessive justifications for power. While our instincts might tell us that those in the opposing political camps are not merely wrong but evil, they are in fact acting sincerely morally, according to different conceptions of morality with a different weighting of moral values. Epistemic humility advises us not to dismiss them too readily, let alone mischaracterize them.
The final motivation for compromise is that of recognizing other voters’ power—the ability to impose one’s will on others. This situation could arise from certain individuals having more power than others, or certain voter segments (shared values, interests, and/or identities) having greater numbers. In the case of corporate elections, shareholders wield power according the number of shares they own. In democratic elections, the result naturally reflects the voter numbers.
Sometimes what matters is not actual voter numbers, but their perception, as in most elections the actual number is fuzzy, and turnout is a big unknown. Under voting systems that do not naturally produce compromise results, voters whose favorite candidate is perceived to be unpopular face a dilemma: 1) vote sincerely for their favorite candidate, who will probably lose, leaving their opinion unrepresented, or 2) vote for a less preferred but more popular candidate.
One may recognize this from game theory as something like an N-player Prisoner's Dilemma, and indeed such voting systems turn voting into a game—making a decision based on what one expects other players to do. Acknowledging their lesser power, minority voters are compelled to choose the second option; in voting theory this is called the compromise strategy.
A voting system that creates this dilemma is said to fail the Favorite Betrayal Criterion. Cardinal voting systems are not vulnerable to it; ordinal (ranked) voting systems are. The choose-one plurality voting system common in the United States is horribly vulnerable to favorite betrayal, which is a regular feature of most US elections. The compromise strategy is often the best one can do when stuck with a bad voting system.
Consensus and Compromise
One often hears consensus defined as something approaching unanimity of opinion, but for large groups this is rare. A more reasonable definition is general support for the decision-making process, and agreement that the final result is the best obtainable, reflecting input from all the voters. A consensus decision will probably be a compromise decision, with concessions from nearly all voters. Group decision-making should not be a game, and there should be no “winners” or “losers”, at least not if the group expects a long-term future together.
Groups whose members are consistently unable to recognize and incorporate the different interests and knowledge of other members, or who consistently find themselves forced to bow to other members’ power, should consider separating into smaller, more harmonious groups.
First, however, they should give compromise-producing group decision-making systems a try, with a clear conception of what compromise really is--a good thing.