This text originally arose as a portion of a side-note in another text. Due to the relative importance of the topic and the only tangential relevance to the other text, I decided to put the contents on a separate page. Through a quirk in my writing process, the original contents remain very “keyword-y”, while further contents occupy most of the page. (While I publish the page as is, I do not view it as finished. The “keyword-y” part should be expanded and there are other sub-topics of interest that currently go unmentioned.)
Below, I deal with the quality of decision-making. I do not rule out that collective decision-making can have a place for other reasons, e.g. that, in some context, someone with a legitimate stake in some decision/project/enterprise/whatnot should have a right to a vote instead of just having some single individual decide over his head. (However, chances are that the best solution to such issues is to minimize decisions that affect e.g. a collective by any decision-maker, be it an individual, a committee, or the collective as a whole. A great many such decisions are simply unnecessary and should have been left to be decided on an entirely individual basis to begin with—to the degree possible, we should all be allowed to decide for ourselves. Note e.g. how many governmental decisions are superfluous or outright harmful, yet apply to everyone, and a below discussion of my “apartment owners association”.)
In the case of politics, concerns like “checks and balances” also must be given due weight, together with the observation that concentration of political power in the hands of the few has great risks and has often ended badly—but mostly for reasons like a wish for personal glory, not because a smaller group of politicians/rulers/whatnot would tend to be worse decision-makers than a larger group. (Whether politicians in any size of group reach a satisfactory average level is another question. Similar concerns might apply in other contexts, e.g. large corporations.)
Committees and other mechanisms for collective decision-making have a very poor track record. Common problems include:
A lessening of individual responsibility and a greater risk of “moral hazard”.
Herd mentality and group think.
The influence of group-internal politics on decisions, including those affecting third parties.
The risk that the group (committee, whatnot) takes on a life of its own and beyond is official purpose.
Corruption among individual group members. (Note that it might be harder or much harder to corrupt the entire group than to corrupt individual decision-makers, but that it is often sufficient to corrupt just a few members, e.g. to tip a 6–4 vote into a 4–6 vote.)
An increased need or undue willingness to compromise. (For the overlapping issue of “consensus for the sake of consensus”, see below.)
The extremely common issue of the dull majority overriding the bright minority (or a charismatic dullard manipulating the majority to vote according to his whims while the bright non-manipulator is left with only his own vote).
To give just a few example of where collective decision-making might lead: At the time of writing, March 2024, Joe Biden is POTUS as a consequence of outright collective decision-making, while, as signs of where collective decision-making could take us, Adam Sandler appears to be the best paid actor in the world, and Taylor Swift the best selling “recording artist”.
Note that the above list contains issues that can lead to poor decisions. Not included are problems like e.g. a greater need for coordination, which are more likely to lead to problems like loss of time and waste of resources than to worse decisions. In many practical situations, however, such issues can be of great importance, as with group work (cf. below).
Based on both my own experiences to date and theoretical considerations by me and others, and in terms of finding a quality decision, I would see collective decision making, decisions made by groups (including the infamous committee), and similar a bad idea much more often than not. (I caution again that other factors than quality might be relevant.)
However, this does not imply that decision makers should not consult others. On the contrary, gaining input from others is often at the core of good decision making, including to compensate for own knowledge and experience deficits in a given domain, to see different perspectives, to have some idea of what others actually want and need (especially important for a political ruler), and similar. Key points include: (a) to keep the ultimate decision in so few hands as possible; (b) to not fall into traps like a belief in the “wisdom of the collective” (let alone the preposterous idea that a democratic decision would be infallible); (c) that the decision maker(s) truly use their own heads. None of these key points speak against consulting others.
Even here, however, it is disputable whether the group should be consulted as a group (as with a focus group giving product feedback), the collective as a collective, etc. It will often (likely: usually) be better to approach group members individually. (The exact nature of such an approach will depend strongly on the group at hand and other varying circumstances. This might be explored further at a later date.) In particular, thoughtful and in-depth feedback from some few tends to be vastly more valuable than “On a scale of ...” questionnaires, and similar simplistic forms of feedback, by every single group member.
From another angle, an issue that often strongly contributes to problems with groups is the usually lower average intelligence, competence, domain knowledge, whatnot level, compared to decision-making by the single best brain. This issue can be alleviated by greater care when selecting the decision-making group, giving groups a greater justification in the decision-making process. This, especially, if the group is kept small. (Also note the risk that some single brain actually put in charge is far from the best, in which case the group can form a counter-weight. The better solution, however, is to keep that single brain well chosen.)
The popularity of juries is not an argument for collective decision-making:
Firstly, popularity is not a proof of quality. (The opposite assumption is, of course, another case of bowing to collective decision-making, if in a more indirect form.)
Secondly, (court-room) juries are not used to achieve a collective decision of e.g. “by a majority vote and the wisdom of crowds, we declare the defendant guilty [innocent]”. The point is rather to have (typically) twelve individuals reach a consensus that the defendant has been proved guilty beyond reasonable doubt—or to reach a consensus that sufficient proof has not been given.
Indeed, chances are that the jury system would work better with less “jury deliberation”, which is a case of collective decision-making, and more individual and independent thinking. A potential way would be to have to jurors deliberate separately until they feel certain of a vote and then to let them vote. If the result is unanimous, we have a result; if not, some additional discussion to exchange reasons, clarify misunderstandings, point to errors of thought, whatnot, in the hope of reaching a consensus is warranted, but this should be kept within limits. If no consensus can be reached, or if more than one or two jurors dissent, past a certain time, then “reasonable doubt” should be presumed.
If twelve jurors cannot agree on a “guilty”, this, in and by it self, is a very strong sign of “reasonable doubt”.
It might well be that one or two jurors are caught in preconceptions, have been bribed, or similar, and some special consideration for this might be necessary, but the current U.S. system of an insistence on consensus and the current handling of “hung juries” is problematic. The U.K., in contrast, appears to have a limited room for a clear majority convicting over one or two dissenters. To consider e.g. a 6–6 vote a “hung jury” instead of proof of “reasonable doubt” is perverse. (Also note a discussion of “consensus for the sake of consensus” below.)
A common “proof” of the superiority of crowds is some simplistic demonstration like a group of individuals estimating the value of something, with the result that the average estimate was superior to most or all individual estimates.
Apart from this being relevant only to certain types of problems, and usually problems that are better solved by (literally or metaphorically) having some individual member of the group grab a ruler and actually measure:
This is a trivial statistical phenomenon, likely to hold almost automatically in a wide range of cases. Likewise, having a group throwing darts at a dart board while blind-folded (but having some general idea of where the board is) is likely to find the average and/or median throw superior to most or all individual throws. (If the average is used, some minor data corrections might be beneficial, e.g. to remove throws that failed to hit the wall. Such corrections are very common in statistical investigations.) It is even conceivable, especially with estimates that involve more guessing than skill, that letting the same individual member of a group estimate repeatedly gives an average estimate that tendentially improves with the number of attempts. (Barring the significant complication that the individual might remember his previous estimates.)
From a more mathematical point of view, assume that each individual makes an estimate that deviates from the true value according to a normal distribution and that the distribution is the same for all individuals. (This is not realistic, but is often somewhat close to the truth, and it does illustrate the principle well.) The expectation value of both each individual estimate and the average estimate will now coincide with the true value, while the standard deviation of the average estimate will be the standard deviation of the individual guess divided by the square-root of the number of guessers. For every quadrupling of the number of individuals, we can expect the average guess to become, in some sense, “half as far away” from the true value.
This is not wisdom—it is statistics.
Even apart from such considerations, it is not unexpected that the average (often better: median) fares better than most individual estimates. If we, for simplicity, look at the median, half or more of all individual estimates will be at or larger than the median and half at or smaller than the median. If the median hits the true value exactly, it obviously “wins” in the sense of being better than most individual estimates. If not, the true value is either smaller or larger and, regardless of which, at least half of all estimates are at least as far from the true value as the median—and the median “wins” again. (A draw could take place, but, except for very small groups, this changes little. Also see a more detailed discussion of a similar issue, including some thoughts on values that “draw” with the median.)
As to the quality of estimates, just as with dart throwing above, note that having a computer spit out some random numbers would yield the same result—the median “wins”. This even if the numbers are not even in the right ballpark. For instance, the median of 1, 3, 4, 11, 12 (i.e. 4) is at least as good an estimate of 100 as three out five entries (1, 3, 4), and loses against only two (11, 12). Ditto, if we change “100” to “‘100 million”.
A related problem is the common obsession with group work in school. Such group work usually ends up as a whole that is lesser than the sum of its potential parts, while showing many examples of poor decision-making, unfair distribution of work and credit, etc.
While most of this would be off-topic, group work can serve as a good illustration of why so many have so naive ideas about collective decision-making. Consider three types of students in a group of X members:
The uninformed and unintelligent, who would be over-challenged by the project at hand, if left to do it themselves. (Even allowing for sufficient additional time to compensate for the loss of manpower.)
These see a task to complete that seems daunting, they have no idea where to begin, and, during work, do not necessarily understand who contributes what. However, they do see that this daunting task was somehow completed by the group.
Yay group work!!!
(The list of types is not exhaustive and others can have a similar reaction for different reasons, e.g. in that someone lazy can often duck work without endangering the completion of the project.)
The midrange students, who might manage the work alone if given sufficient time, but not with a particularly high quality. Their main take away might be “I would have needed X days to get this done, but the group got it done in a single day!”.
Yay group work!
(Factoring in overhead, the group would likely need longer than a single day, but let us be generous.)
While they are likely aware of the differences in contributions, e.g that the first type does not contribute in proportion to number, they might have a skewed view in other regards. For instance, they might value leg work as highly as head work or measure contributions based on time put in, not in quality or quantity of result or ability to make good decisions. (As to quantity: A faster worker might get as much done in an hour as a slower in two hours. Should we then value the one “fast” hour as equal to one “slow” hour or to two?)
Worse, they might often be annoyed at the students of the third type (cf. below), because these believe (usually, correctly) that they know better than the wisdom of the crowd/group, have higher quality standards (“are picky”), prefer head work over leg work, whatnot.
The top students, who could do a better job than the group and in a shorter time. (Measured in e.g. “man-hours”, not necessarily “clock-hours”.)
These see group work for what it is and are frustrated by how much work they have to put in relative the freeloaders in the group, by how little the others seem to care for obvious quality problems, by being overruled by the majority, even when the majority supports nonsense, by how much time is lost on coordination and debates, by losses to “forming/storming/norming”, etc.
Boooooh group work!!!
Interestingly, one of the greatest problems with group work is exactly collective decision-making, in that we might have a handful of students supposed to “decide together” so that they will “learn teamwork” because “teamwork is so important in the office”. In the actual office, however, there is typically a designated team/group/project/whatnot leader who makes core decisions—and “decide together” might be an outright counterproductive approach for working well in a real team in the office. Likewise, a typical sports team might have a team captain, a coach, a team manager, and often some other roles too, that make many of the decisions—and especially the bigger decisions. A sports team does not usually get together in a huddle to jointly decide what to do next. (More common uses of huddles include communication of an already made decision and mutual “pepping”.)
Other differences exist, e.g. that group work in school is performed in ever changing groups that must ever again begin over with “forming/storming/norming”, while teams in the office are usually stable for much longer and spend a far higher proportion of their time “performing”. At extremes, we compare hours with years.
At work, collective decisions have very often gone a different way from what I suggested, because the collective, say, ignored dangers that I pointed to, did not understand advantages and disadvantages that I mentioned, shrugged off my long-term concerns in favor of the here and now—and I have usually (!) been proved right by time. This especially in a Scrum context, and too much collective decision-making by too poor heads might be the single greatest weakness of Scrum (of which I am otherwise in favor). A particularly interesting event took place one of the first few days in a new project in a new company:
We were playing Planning pokerw to estimate some new tasks in “story points”.
See the link for information on terminology, if needed. Note that the scale used deliberately jumps in a near-Fibonacci sequence, 1, 2, 3, 5, 8, 13, etc.
The exact numbers and the exact order of this-and-that given below might be off due to the many years since the event. This is not harmful to the illustration of principle.
I offered an 8 for a certain task, the others were stuck at a mixture of 3 and 5.
After some discussion and one or two renewed votes, I was still at 8 and the others unified at 3.
The others now insisted that I change my vote, because “Scrum demands a consensus!”. I countered that I could with equal justification claim that “Scrum demands that we call them as we see them!” and offered a quite reasonable way out—as this was my first time in the new team, we should simply disregard my vote. (This with a solid rationale: Story points can vary from team to team and even from time to time within the same team, implying that I might simply have been used to a different “size” for a story point from my previous project. However, as time showed, the differences were quite small and my 8 was an 8 even by the standards of the new team.) The offer was refused—every vote must be counted and the vote must be unanimous. (That this was a USSR-style thinking seemed to escape them: You must vote and you must vote according to the party line! Of course, such an approach is entirely contrary to the purpose of planning poker.)
I re-iterated that I would not make a dishonest call and repeated my offer, which was now very begrudgingly accepted, with sighs and a clear demonstration that I was considered “not a team player”.
Once the task was actually implemented, my 8 turned out to be the right answer. Indeed, if anything, it might have been a 9 or a 10, had the scale been more fine grained.
I did not get belated recognition for having stuck to the right answer—but I did repeatedly note that this was one of the least competent teams that I had ever worked in.
Apart from general illustration, two particular and recurring-in-my-experiences points stick out:
Firstly, a drive for “consensus for the sake of consensus”, as if consensus would be more important than getting the right answer. It is true that planning poker is based on a consensus, but this consensus should arise through a sufficiently shared understanding that the participants can genuinely agree—and genuine agreement and “consensus for the sake of consensus” are very different things. (And with the near-Fibonacci sequence used, this is usually easy to achieve or achievable with only minimal compromise, as e.g. going from a “weak” 8 to a “strong” 5, or vice versa, is a smaller jump/compromise in understanding than what it might seem to the beginner.)
While a genuine consensus is better than a “majority wins”, “majority wins” is usually better than a “consensus for the sake of consensus”. Exceptions include e.g. an acquitting jury, where a minority favoring acquittal should trump a majority in favor of “guilty” (cf. above).
More generally, groups tend to have issues with poor approaches to decision-making. Planning poker done right can work well, in part through drawing on the statistical mechanism discussed above, and could be seen as an attempt to overcome the risk of a poor approach to making estimates. (At an extreme, “estimates” might involve a project manager saying “We only have 2 man-days left in the time budget; ergo, I put it down as 2 man-days. Anymore and the critical path goes beyond the deadline.”.) Here planning poker was not done right—and the bigger point of the anecdote is not that “I was better at estimation than the team!!!”, which could have been sheer coincidence, but to illustrate the type of problems that can be found in collective decision-making.
Secondly, if partially off-topic, my understanding of “team player” is often at odds with that of others: I see it as putting the well of the team ahead of my own (within reasonable limits); many others seem to see is as a matter of blind compliance, including never questioning “the way that things are done” and going for “consensus for the sake of consensus”.
Another good illustration of collective decision-making and whatnot is the “apartment owners association” (“Eigentümergemeinschaft”) of my apartment building and its yearly meeting. There are nine members, one for each apartment. Excepting myself and with reservations for changes over time and a necessarily superficial impression of some members:
About half of the members seem to prefer to just give the building administration a power of attorney to vote on their behalf, possibly with some specific instructions on specific questions. (The power-of-attorney form distributed by the building administration contains phrasing to the effect that “where I have not given explicit instructions, the administration can use my vote however it wants”.)
The other half consists of average persons with no particular qualifications, intelligence, and whatnot. None of them is someone that I, e.g., would hire for a position of responsibility. (The same is likely to apply to the first half, too, but this is harder to judge and of secondary importance due to the power-of-attorney issue.)
All, or almost all, seem to be content to vote for the suggestions of the administration in a blanket manner, without true understanding, true debate, true questioning of motives, true whatnot—and the administration has repeatedly demonstrated a corresponding attitude of “We make the decisions. The purpose of the yearly meeting is to rubberstamp them as a pure formality.”. (Keep in mind that we are the owners, while the administration is merely hired by us to provide services.)
A particular idiocy, as an outgrowth of this collective decision-making, is cable TV: Despite the ever lower relevance of TV, in general, and despite the many channels available “terrestrially”, the collective decided to renew an existing cable-TV agreement, through which every apartment, regardless of use of TV, would be bound to pay twenty-something Euro per month (!) to a cable company. This without any cable actually being delivered, as this portion only ensured that the apartment was connected to the network—each and every apartment that actually used cable TV would have to pay a similar and separate amount per month! The collective bound us for 10 (!) years to this nonsense, despite the foreseeable massive further reduction of the relevance of TV during those 10 years. In my case, as a non-viewer, that is some 3 thousand (!) Euro down the drain for nothing. (And on top of the outrageous governmental charges to finance the third-rate public TV stations, which comes to a similar amount per month—and which also gives me nothing in return.) Importantly, this decision was made based on entirely insufficient information, what amounted to an advertising flyer provided by the cable company.
If solid information had been present, mere differences in priorities could have provided a partial explanation. With the available information, the decision could never have been more than “right for a poor reason”. (As is, it was wrong, regardless of the level of information.)
I managed to get hold of true information from the very begrudging administration and the view created by the flyer was highly misleading. While this took place too long ago for me to remember the details, I note e.g. that the flyer promised fixed prices through the 10 years (which traded fixed prices for a fixed length of contract), while the true information contained reservations that allowed the cable company to raise prices anyway (while we were still bound to the 10 years). A particular annoyance was that the administration obliged it self contractually to help the cable provider spam us, where it should have been the administrations duty to act in our, its employers, favor.
Looking at the situation in a bigger picture, this system is entirely perverse. We, as the apartment owners, have an asset that the cable company (!) should pay for, namely access to the inhabitants of the apartments. The cable company should pay us to gain that access and then get whatever profit it can from the inhabitants.
The cause of this idiocy is likely two-fold (not counting the irrationality and stupidity of all too many apartment owners Germany-wide): Firstly, many of the owners are (like I am) also inhabitants and might (unlike I do) fail to see the “we have an asset” aspect and only see that “in order to get cable in my apartment I have to pay this”. This likely (again, unlike I do) while failing to do the math of “paying twice” (or thrice, if the governmental fees are counted in). Secondly, among the “own to rent” owners, there is little downside to this, as the cost for the cable company can (according to prior legal cases) be passed on to the renters in full—and regardless of whether the renters actually use the services of the cable company, which, again, requires a second fee from the renters to the cable company.
In addition, there are rumors (but I have, to date, not seen proof) that building administrations receive outright kickbacks from the cable companies for keeping apartment buildings contracted.
From another angle, we have two issues of others making decisions that we all should make for ourselves (even if we ignore issues like who has an asset and who truly should pay whom): Firstly, we should all, as apartment owners, have the right to accept or reject the cable company individually. The only decision that might be relevant for the collective is what company should be allowed to offer services, because there might be technical restrictions that prevent multiple companies from offering services simultaneously. (But note how the current situation removes all incentives to improve on this and how e.g. different households can have contracts with different energy providers, even while in the same house.) Secondly, renters should not be forced to accept landlord decisions on such issues, and certainly not be forced to pay for unwanted services. There are some services that more-or-less have to be unified, e.g. in that the stairwells are cleaned in a unified manner and the costs divided without regard for who goes in and out of the house how often, but no such “have to” is present for cable. Similarly, all households have separate contracts with separate providers for electricity, water, Internet, and any similar service that they use—cable TV is the sole exception.
In a bigger picture, the situation is a microcosm paralleling real world governments, including that the administration tries to hinder and manipulate democratic processes, push the decisions to its own advantage, etc. For instance, I have made repeated suggestions for issues to bring to discussion—including that we all be provided with actual information, not advertising flyers, before decisions. These have all been silently ignored by the administration (which is legally in charge of setting the meeting agenda), meaning that they have been off the table with neither discussion nor vote. For instance, the location of the meeting has invariably been so far from the building that it has been hard to reach. For my last attempt, I checked addresses around two hours before the meeting, found that I would need about an hour (!) to get to the location, and would need to travel a significant distance by bus to get there. (I gave up.) This despite the option of using some part of the house for the meetings and despite their being several restaurants and whatnots, that would have served perfectly well, within a few hundred meters of the building. For instance, there has ever again been agenda items to the one-sided favor of the administration at our cost, e.g. that the meeting be held earlier in the evening (greater convenience for the administration, less so for owners who work and/or commute; also remember the “who pays whom” factor).
Similar errors to the flawed statistical argument around crowd estimates occur elsewhere.
For instance, I have seen the occasional claim that “Science shows that average faces are the most beautiful!”, with the twist that it can create the impression in the careless (or wishful) reader that having an “average” face in the sense of e.g. a 1–10 scale would be better than having a higher ranked face. This impression, however, shows a similar issue with understanding to the “wisdom of crowds” idea: In the former, “average” is, matching everyday use, taken to be someone “average looking” in the sense of “everyday looking”, “regular, non-model, woman”, or some such, while the implication of the claim refers to average in a statistical sense (cf. below). In the latter, some idea of crowds being better than individuals at estimation is implied, while the true explanation is a trivial statistical phenomenon (cf. above).
Here, I am uncertain whether it is one single investigation that has been re-trumpeted on several occasions or several different investigations using the same idea, but the idea is dubious. (I am also uncertain whether this dealt with only women’s faces or whether the same was done with men’s faces and, if so, with what result. Most of the below would apply to men too, but some details might vary, e.g. relating to age and skin quality.)
What happened is that a great number of faces were taken and an “average” face was constructed by image processing. There are now at least three mechanisms, similar to the above, that play in:
Various skin blemishes are averaged out, leading to a “smoother” skin than all (or almost all) the original images. Such skin, however, is naturally viewed as beautiful.
Various facial asymmetries are averaged out, leading to a more symmetrical face than all (or almost all) the original images. Symmetrical faces, however, are naturally viewed as beautiful.
Both human perception and human variation is likely to be centered around some “average” in various features. (Of which there are likely dozens or more.) Just as with the above estimates, the average of the faces will be closer or much closer to these “averages” than all (or almost all) faces in any individual feature—and will certainly be so over the sum of all features. Human perception, on the other hand, is naturally geared towards these averages, at least to the degree that something “too large” or “too small” is viewed as odd.
(With the caveat that the underlying “averages” might differ based on ethnicity, sex, age, or similar. For instance, men tend to have facial features that are more pronounced than do women.)
In all three cases, we can suspect an evolutionary mechanism related to health, e.g. in that someone with a smoother skin is likely to be healthier, all other factors equal, and therefore be a better prospect as a mate (and, likely, some other roles; the too unhealthy might need to be outright avoided, as some contagious disease could be involved). In the specific case of skin, there is also likely to be a greater perception of youth, because skin tends to degenerate over time, and younger women are tendentially, within reasonable limits, viewed as more attractive.
The first two are not compatible with the claim about “average faces” at all, but reflect a pure statistical issue (and/or methodological issue): taking averages leads to smoother skin and greater symmetry—and it is the skin quality and symmetry that gives a positive impression.
The third does give some justification for the overall claim of average being attractive, but (a) it is only one of three factors, (b) the result is still a matter of statistics, (c) it does not match the impression of that careless reader.
However, outside of and partially contrary to such investigations, the careless reader could have a point for other reasons: I do, myself, often find a more everyday or less perfect face more interesting, simply because the “too” perfect, “too” smooth, “too” symmetrical, “too” whatnot, can be bland and boring. (Similarly, compare the sterile and boring rooms sometimes depicted in e.g. house and furniture adds with a room that someone actually lives in; or compare a musical note as a pure sinus tone generated by a computer with the same note played on a musical instrument.) This view, however, is not compatible with the view of the investigations, as the constructed average faces would be “too” on all counts.
A common problem with collective decision-making is deliberate abuse, especially for the purpose of avoiding individual responsibility. For instance, a “true” decision-maker might hide behind a committee with a majority of “yes men”, the easily manipulated, whatnot, that almost always votes in the “right” way—as if the decision-maker had also nominally been the sole voice. When things go sour, this decision-maker might then still invoke some variation of “It was not my fault! We decided together!”. Worse, the decision-maker can claim, regardless of success or failure, that “I was against it, but I had to bow to the majority!”.
Similar ideas can apply in other areas.
For instance, a poor grade on a group project in school (also cf. above) might then be “Not my fault! The others screwed it up!”. (Useful e.g. to lie to oneself, to defend oneself against criticism from a parent, to bargain for a better final grade with a teacher.)
For instance, I have long suspected that many EU members and/or the parties currently in charge of EU members deliberate push for the EU to make unpopular decisions so that they can say “The EU made us do it!”, without acknowledging that the EU merely gave them what they wanted in the first place.
But is this not a good work-around for groups? This way, there often will be just one decision-maker!
No:
Firstly, there is a world of difference between (a) a sole decision-maker (or a sufficiently small group of decision-makers) who is also known as such and takes later responsibility as such, and (b) a sole decision-maker who is protected from the need to take responsibility.
Secondly, such hidden decision-makers are not usually chosen for their brilliant decision-making, which removes a central point of having fewer decision-makers, but gain power from e.g. their skills at manipulation.
Thirdly, as noted at the beginning of this page, there might be other reasons than quality of decision (etc.) that must be given due consideration, e.g. “checks and balances”, but could be circumvented by such mechanisms of hiding.
A great potential complication is decisions deliberately being moved from a small group of competent individuals, or a single competent individual, to a larger group of less competent individuals, which is easier to manipulate into the decisions that some string pullers want. The decisions tendentially grow worse, the string pullers are protected from the repercussions of poor decisions, and there might even be an option to push an alleged quality of the decisions through the larger group. Many of the naive seem to take a larger group as something automatically beneficial, up to the point that a democratic or pseudo-democratic decision in national politics is hallowed.
National politics, however, does not only suffer from many poor decisions, but is often very far from something truly democratic. Note e.g. the many votes that result from deal-making and the problems with prescribed block voting—either you vote according to the one-man decision that constitutes the party line or the party whip will be the ton of bricks that ends you. Then we have issues, in multi-party systems, like a mere plurality party being supported by nominal majority decisions.
After the original publication, I encountered a speech by Antonin Scalia (“Scalia Speaks”, heading “Dissents”) that well illustrates issues around the value of dissent and the dangers of consensus for the sake of consensus in the context of the U.S. court system. Much of what he says is valid in a much wider range of contexts, however.
While I refer the interested reader to the actual speech, a few points of particular general importance, abstracted from his discussion:
Acknowledged dissent is good for the market-place-of-ideas and, in particular, makes it easier to achieve changes at a later date
A focus on reasoning and persuasion over consensus leads to better decisions, be it because the majority positions changes over the course of the discussion, because details are improved, or because consequences are better understood in advance.
Such focus often leads to a better understanding of the matter at hand and better future decision-making. (Even should the eventual majority position remain the same in the case immediately at hand.)
Dissent replaced with artificial consensus can give outsiders the wrong idea, cause them to make flawed decisions based on a perceived consensus, set flawed precedents for others, and similar.
With an eye at my own experiences from the office, problems like reasoning that “We will do X, because we have always done X!” and “X has always been good enough for everyone else, so why should it not be good enough for you?” often depend on prior consensus (whether genuine or faked). Allowing louder dissent, preferring genuine majority decisions over fake consensus, etc., makes such reasoning less viable and increases the chance of future improvements. (Such reasoning seems to be particularly common with weak thinkers and/or those who have only weak arguments to support their opinions.)
Since the original writing, I have done some reading on topics like AI and machine learning. Here, a type of decision-making of a collective nature is often, and with success, used in the form of “ensemble methods” like “bagging”, which has a considerable similarity with the above statistical averaging. This can e.g. be done by taking several trained models and having them “vote” on some matter, but where the matter is something more complex than, say, what numerical value some specific whatnot will have, and can involve whether a certain image does or does not contain a certain type object.
To avoid confusion around the next paragraph: Formulations involving “model” relate to machine learning and not e.g. fashion magazines. (And, yes, it took me quite a bit of self-control to not attempt a few puns.)
My readings in this specific area are as yet very superficial (I recommend a grain of salt to the reader), but the underlying idea seems in part to be a matter of statistics similar to human estimates, in part a matter of removal of blemishes similar (in a metaphorical sense) to averaging of faces. Where a single face might contain a mole, an averaging over increasingly more faces will make this mole gradually disappear and an averaging over several trained models can make a weakness in one of the models disappear in a similar manner. (With the likely difference that a vote results in a more abrupt disappearance.) To take an analog example in human terms, say that we have several beauty judges determining whether a young Cindy Crawford should be awarded a modeling contract. One of them is hung up on her mole and votes “no”; the others are not and vote “yes”; the overall result is a “yes” and history bears the decision out.
This example is fictitious. I have no insight into actual events.
Note that beauty is ultimately subjective, but that even the subjective can have objective effects, e.g. in that a face subjectively deemed beautiful by enough magazine buyers and whatnots can bring about very objective flows of money—or launch a thousand ships.
For issues where the subjective opinions, tastes, whatnot, of the masses are relevant, a less intelligent judge can actually have an advantage, because he is more likely to share these opinions/tastes/whatnot.
In a next step, we have the question whether such ensemble methods might be an argument in favor of human decision-making in a collective, even beyond e.g. matters of statistics. The answer is “yes” in as far as it can help with removing errors like in the preceding paragraph. It might even be, when discussion is allowed, that a human recognizes an error in his own reasoning or decision-making and is willing and capable to adapt. However, caveats apply when the same idea is used more generally, including that majority voting (or other reliance upon a majority) can be more beneficial for subjective issues than objective issues (cf. above side-note), that what works in a straight vote (where everyone makes his own decision) might not work when pre-vote influence is allowed (be it among voters or from some third party upon voters), that a majority vote between a minority of the intelligent and knowledgable and a majority of the dumb and ignorant is unlikely to work well, and similar. Notably, while trained models can be fallible, they are normally all experts of a sort in terms of the decisions and whatnots for which they have been trained and they do not underlie the types of weaknesses that humans so often do in terms of collective decision-making.
That said, I do not argue against the possibility that a group of sufficiently good human judges who approach a decision sufficiently rationally can achieve better results than an individual judge. The issue is the relative rarity of such scenarios among humans and the relative probability that there exist some individual judge in the group who is sufficiently superior in judgment that he beats the group. A particular complication is that the case of someone (as above) recognizing an error in his own reasoning or decision-making is likely to be rarer than cases where someone (literally or metaphorically) changes his vote in order to conform, to not rock the boat, to avoid leaving a negative impression on others, as a result of an unwarranted suspicion of an own error (note e.g. the Asch conformity experiments), or similar—up to and including cases of someone agreeing or pretending to agree for the sake of consensus.
The mistrial verdict (with reservations for terminology) in the trial of Lindsay Clancy for the killing of her three children has been major news in September 2026. Here, a majority of the jury wanted to acquit based on claims around post-partum depression/psychosis and whatnot, while a holdout juror saw her as guilty, in a reversal of the “Twelve Angry Men” scenario so often used in fiction. (That she killed the children is beyond controversy. The issue was whether she could be aquitted on some other basis, e.g. mental illness.)
The trial and the underlying case, as such, are of little relevance to this page. However, the jury situation is interesting on several counts. Note e.g. earlier words on specifically juries and my own experiences with an incompetent team playing “planning poker”.
In later public criticism of the holdout and (to the degree that reports around the internal workings of the jury can be trusted) his situation during the trial, there are recurring themes like it being wrong for him to hold the rest of the jury back, that he wasted everyone’s time through not caving, that he had a duty to join the consensus, that a nominal requirement of a jury consensus should be replaced with a mere majority decision or an almost consensus through dissenters adjusting their votes to match the majority, whatnot. (With variations in detail depending on source/speaker/whatnot.) It also appears that he has been both pressured and misrepresented.
Contrasting earlier claims by others with later claims by the holdout, misrepresentation seems to be present in at least (!) two areas:
Firstly, that he would have acknowledged “reasonable doubt” yet failed to vote for “not guilty”, while he denies both having and acknowledging such doubts.
Secondly, that he would have been a truly lone juror defying eleven solid acquitters. His version is that (a) the original division of the jury was very far from an 11-1, (b) even towards the end it was a 10-2 until a second dissenting juror caved under pressure.
A potential third, but where the claims that I have seen are vaguer and/or my memory too imperfect, is that he was accused of being driven by an agenda (especially, in conjuncture with an alleged acknowledgment of “reasonable doubt” per “Firstly”), while there are strong signs that, on the contrary, several of the acquitters had an agenda of pushing some political or other “women’s issues” angle, wanted to use the trial as a means to “start a conversation” about post-partum depression, or similar. While I will not enter into such matters below, it can pay to keep in mind that such an agenda could result in many arguments being raised in bad faith instead of out of e.g. ignorance or personal convenience, say, by invoking the below “Be reasonable! If you just change your vote, we can all end this and go home!” in order to gain an acquittal (intellectually dishonest) instead of in order to go home (lazy and unprofessional).
If the jury system (ditto, planning poker) is to work, however, we must have a sufficient degree of freedom of opinion/conscience/whatnot—a juror who believes in the guilt or innocence of the accused must be allowed to hold and express that belief. (Including through voting accordingly.) What the other jurors believe must not be a consideration in this regard, nor must they be allowed to exert pressure of e.g. a “Be reasonable! With 11 to 1, you must be wrong!”, “Be reasonable! If you just change your vote, we can all end this and go home!”, or a “Be reasonable! If you do not change your vote, the entire trial will be for nought—all that wasted time and resources!” kind. (And imagine how differently “Twelve Angry Men” would have played out if the holdout had caved to “guilty” based on similar pressure.) Likewise, each juror must be allowed to originally form his own opinion based on his own evaluation of the proof and arguments presented by the prosecution and the defense.
Nothing above is contrary to an exchange of factual arguments between jurors, e.g. to point out errors in reasoning by another juror.
However, even here some care might be necessary, as arguments going beyond those by the prosecution and the defense could risk distortions, create spurious “legal theories”, or otherwise be problematic. For instance, an “if the glove does not fit, you must acquit” (with reservations for exact phrasing) was bad enough when used as argumentation-by-slogan by the defense in the infamous O.J. trial—but what if one of the jurors had come up with this slogan and pestered the rest of the jury with it?
Pressure on a “[...] the entire trial will be for nought [...]” basis is particularly dangerous as (a) seemingly juxtaposing the interests of a lone juror with a greater good in combination with shaming or “guilting”, (b) implying that it is more important to have a verdict than to have the correct verdict, (c) putting matters of “process economy” above justice, e.g. by avoiding a retrial.
To “seemingly”: In reality, it is the interests of the accused and the public that should be considered—here, in that the public has an interest in criminals being punished and that they not be acquitted for convenience (as opposed to presence of reasonable doubt). In reverse situations, this can become the more dire, as someone might then be convicted (not acquitted) for convenience.
To “process economy”: I am unaware of corresponding English terminology and traditions, but the German system puts extreme focus on “Prozessökonomie”, to the point that some seem to consider it more important to save money for and within the justice system than to get actual justice. Likewise, it is not enough that a crime has been committed for the police to involve it self—there must be a sufficient public interest in the involvement. (And guess who decides whether such a public interest is present...)
If something else than a unanimous verdict should be allowed (and, cf. above, I am very open to this, especially, for aquittals) it must be by some formal mechanism—not by pressure, manipulation, trickery, or whatever else might apply to the trial at hand. Such a formal mechanism might then allow that an 11–1 vote is as legally conclusive as a 12–0 one or that some minimum number of acquitting jurors is seen as proof of reasonable doubt.
However, some care must be taken not to just shift the problem. If an 11–1 vote is viewed as conclusive, the same problem with pressure towards dissenters could be repeated after a 10–2 vote. (“Be reasonable [etc.]”)
Likewise, the issues of the Clancy trial are strong signs that the type of more individual jury deliberation that I suggest above would be beneficial.
For the better part, the above is written without regard to the sex of the involved parties. However, there are strong signs that women are particularly problematic, e.g. through having that consensus-for-the-sake-of-consensus mentality more often, being more likely to vote based on personal sympathies (be it with those affected by a decision or with those arguing for a particular decision), or being more likely to prioritize the approval of others above standing up for a genuine belief.
Indeed, the aforementioned “planning poker” incident took place in a team where both the team lead and the Scrum master were women, and the main conflict was between me and them—not me and the rest of the team. Indeed, the clear majority (8–4?) of the jurors in the Clancy trial were women, while the holdout was a man. Indeed, the representative for the building administration and several notable members of the apartment owners association above are women. Indeed, those who favored group work in school during my own classes were (to the best of my recollection) disproportionately girls. Etc.
With group work other “tendentially female” factors might play in, notably, a preference for a mode of work that allows more interactions with others and lesser confidence in being able to complete a task on one’s own. Whether such factors are a problem in the manner that e.g. a consensus-for-the-sake-of-consensus mentality is, I leave unstated, but they certainly increase the risk that the mode of work is chosen for a poor reason (good reasons relate to efficiency, effectiveness, correctness, whatnot) and that the mode of work leads to collective decision-making. In more adult context, similar issues can make the use of e.g. committees more likely.
I would also not rule out that the pushing of group work in school has a strong correlation with the predominance of female teachers as decision makers. (It is an interesting twist that the choice of group work was usually dictated by the teacher, while one of the alleged purposes of group work would be to learn how to make joint decisions...)
Compromise is a matter often overlapping with one or both of collective decision making and consensus-for-the-sake-of-consensus, used as a tool for one or both of them, and/or something that is often undeservedly seen as a good thing in a similar manner.
Now, compromise can be a good thing, e.g. when two parties have conflicting interests and manage to find a mutually acceptable solution that prevents a conflict that would do them more harm than they stood to gain from winning that conflict (which, of course, only one of them would).
However, even here, it will often be less a matter of compromise and more of e.g. creative problem solving, of gaining a better understanding of the underlying wants of the parties (not the superficial wants), of differing between “musts” and “nice-to-haves”, and similar.
The famous (but presumably fictional) case of the two boys who both wanted the one orange available is not an example of compromise but of the preceding paragraph. (Superficially, both wanted the orange. Looking deeper, however, one wanted the peel and the other the underlying fruit, making the orange divisible in such a manner that both were satisfied.)
Often, however, it is a bad thing—especially, when decision makers, be they collective or individual, are more set on “keeping everyone happy” or “getting enough buy-in” than on making a good decision. Problems include that compromise solutions are often suboptimal, try to do too much, do too little where they should do something, do things in a less focused manner, are more expensive and/or give less “bang for the buck”, whatnot, with variations of the “too many chefs” theme posing particular dangers. A frequent issue is that the compromise solution (if e.g. something developed) is mediocre at two things instead of excellent at one thing. Being mediocre at two things might or might not be preferable to excellence at one thing but, importantly, decision makers often fail to make a proper and conscious trade-off. In German politics, I have often seen a change of rules/taxes/hand-outs/whatnot being complemented with an opposing change to reduce the effects of the original change in order to avoid protests, which then simultaneously reduces the intended effects and makes the overall regulatory system the more complicated.
Whether various examples are compromises in a stricter sense or non-compromises that show similar problems can depend on the details of the case at hand. A true compromise involves the parties reaching an own agreement, with or without mediation, while a half-and-half solution imposed by a decision maker in the hope of keeping both parties happy (as opposed to making a decision between two unadulterated positions) is not a true compromise but merely shares similar problems. (Also note the below case of a meddling mother.) Cases in between might or might not be compromises to some level of strictness.
Also note that even true compromises and the underlying conflicts can have different natures, e.g. in that two boys fighting over an orange or two countries over an island can be a very different thing from two software developers arguing for two different solutions to a software problem. Notably, the former cases see parties that have an eye at their own self-interest and often have high stakes, while the latter aim at the best of e.g. their shared employer or the users of the software with low (personal) stakes. Consider e.g. the next paragraph of the main text and how the latter type of case is more likely to be relevant than the former to that paragraph; while the opposite applies to the second next paragraph.
In many such cases, it is better for the parties involved to not compromise and to force a decision between two (or more) uncompromised alternatives, the better of which is likely to be superior to any compromise solution. (Often, both resp. more than one are.) This is, of course, often contingent on some peaceful and timely means of resolution being available (e.g. a sole decision maker who picks among the proposed solutions or a majority vote; and note that seeking an outright consensus can be just as time consuming as a more overt conflict.)
Even a prolonged conflict can be a lesser evil in many circumstances, because points of game theory have to be considered: If the one side does not put up enough of a fight, the other side might return again and again, looking for further wins (arguably, off topic) or compromises and “compromises” that accumulate to a whole-sale victory down the line.
The last, especially, when the suggested compromise consist in someone giving up something that he already has, to which he has a legal or ethical right, or similar, to someone who merely wants it, as might be the case with many Leftist policies or politics and political bargaining in general. (Consider, with an eye at California’s far-Left nonsense in 2026, a hypothetical “compromise” of “only” putting up a 2.5 percent tax on billionaires instead of the suggested 5 percent. The more so, if it is followed by a later call for a second compromise at 3.75 percent—lather, rinse, repeat.)
As is often the case with similar discussions, childhood scenarios can provide good examples, especially at the hands of mothers with an underdeveloped view of ethics, right and wrong, personal property, and similar. Consider, in a variation of the two boys and the orange, that the boys had each been given an orange the day before, that the one boy had eaten his and thrown away the peel, that the other had kept his, that the orange-less boy now insisted that he be given the remaining orange, and that mother dearest “compromised” by cutting the orange in two—thereby not only committing an injustice but also, likely, thwarting the intents of the boys per the original version.
(Depending on details of intended use, e.g. whether a connected peel was needed, the boys might be able to rectify the situation by adding some Capitalism to the preceding Socialism and having the one exchange his half of the peel for the other’s half of the underlying fruit. This is an interesting demonstration of how Capitalism, free markets, and/or free enterprise, depending on details and the definitions used, can resolve many matters in a manner that achieves what compromise is supposed to, but all too often does not, achieve, e.g. by leading to a more efficient use of resources or mutually beneficial trade-offs. However, even such an exchange does not alter the fundamental unfairness of the original “compromise”, which unduly favored the boy who ate his orange and disfavored the one who kept his. And, yes, Leftism often has an element of “eat my orange and keep it too”, at the cost of others.)
This also illustrates a common problem with alleged compromises, namely, that they are not voluntary agreements between the parties at hand, but something imposed from above. Such an imposition tends to go directly contrary to the alleged main benefit of compromises, that both parties be content, and leave both discontent instead, which increases the risk of future and further conflict—unless the “compromise” is sufficiently one-sided that only one party loses out, making the word the less applicable, and “only” leaving one side discontent. In all cases, it is not only the risk of conflict between the parties that remains or increases, we also have to consider the risk of conflict between one or both parties and the imposer of the “compromise”, who might be resented, viewed as an enemy, or similar, by one or both parties. (Yes, there is often a parallel dictate that the parties be happy about the imposition, but such a dictate does not change the underlying, and often very reasonable, negative reactions.)
There is a fair bit more to be said on the topic, but I save that for a dedicated page, if I ever get around to it, as the connection with matters like collective decision making becomes more tenuous the further the matter is pursued.
The following is an automatically generated list of other pages linking to this one. These may or may not contain further content relevant to this topic.
One head is better than eight legs
How not to do Scrum
Thoughts on help
Sitemap