ad8e

Here's a full list of every page I saw (no selection bias!) and my evaluations.

Anchoring heuristic: this category is not accurate. Anchoring takes two forms: as a priming effect, or as insufficient adjustment from a starting point. Instead, this category lists "Confirmation bias, Echo chamber, Effort justification bias, Escalation of commitment bias, Hindsight bias, Illusion of transparency, Self-fulfilling prophecy". None of these relate to anchoring either directly or through mechanisms. The omission of anchoring, a major effect, is glaring.

Automation bias: the effect as described in the article is not even correct. There is an automation bias, but there is also the opposite anti-automation bias, where humans unfairly disregard the opinions of machines in some contexts, such as algorithmic recipe recommendations. Their CDS example is poorly chosen and barely illustrates the subject. There is a better example from the literature, where humans accept the results of a blatantly wrong calculator over their own estimates. In addition, automation bias is frequently justified even when compared to human "rational" thinking, as summarized in Thinking, Fast and Slow, Ch 21.

I recall this quote from Gelman: "Duncan notes that many common sayings contradict each other. For example, The early bird catches the worm, but The early worm is eaten by the bird." This page on automation bias is no better than one of two contradictory sayings.

I like the three meters, of literature, impact, and replication. Their existence is a well-thought and marvelous insight on science communication. It's a giant improvement over resources from 10 years ago, when such meters were barely considered by psychologists, much less communicators.

I like that they cited references to research papers. I like that they describe how the experiments measure the effect. These are major advantages over comparable websites.

Representativeness heuristic: this category is not accurate. Base rate and conjunction fit. The rest do not.

Base rate neglect: the explanation is quite bad. Kahneman's theory is much more careful and requires huge contortion in "just-so" stories to fit experiments, about how statistical base rates are not always statistical. (I believe his contortions are correct.) But this page doesn't even attempt to describe what qualifies as a "base rate".

The Trump example is a loose fit and the test example is so vague as to be meaningless. The examples in the literature, with criminal identification and test positivity rates, have better writing.

Conjunction fallacy: don't pick that famous Linda example if you have to explain all its linguistic caveats. Your explanations are not convincing, just assertions which the reader can't tell the truth of. (And there are better explanations, like replication under single evaluation vs joint evaluation, or replications with clarifying language, which are not given.) Nevertheless, this page is broadly correct. It is over-specific, focusing on the direct violation of conjunction rather than the representativeness process that underlies the probability estimation, but perhaps the specificity is justified if one wants to hew to the literature.

About->Our team: oh my god, the authors are psychology PhD students. I think psychology needs more academic interest from students, so that graduate programs can have stricter filtering. This website does not give hope that psychology will cast off its bad reputation soon.

  Rebekah Neumann roped in WeWork staff to lobby phone companies to remove a 5G cellular antenna across from their Gramercy Park home.
  Rebekah was scared of a 5G cellular antenna next to their apartment — afraid that electromagnetism from it could cause cancer.
  First, WeWork’s CFO Artie Minson, who had previously worked at Time Warner Cable, was asked to persuade Verizon or Sprint to find a new home for the antenna — despite him busily preparing WeWork for an IPO. Minson didn’t get far, so others in the organization were roped in.
  Finally, it landed with Maria Comella, WeWork’s top public affairs and policy executive. She had been chief of staff to Gov. Cuomo and an aide to Chris Christie. But in the spring of 2018, she was tasked with calling up phone companies and pleading for her boss and his wife: Please move this antenna.
  “We understand that the Neumanns got to a point where they found a way to buy out the cell carrier’s lease and would have been able to have it taken down,” co-author Eliot Brown told The Post. “But that happened right as things were imploding [at WeWork]. To the best of our knowledge, it did not come down.”

Current best method for solving ODEs is PI control. (At least, when I checked it a few years ago.)

RCV in the US: "2 states, 1 county, 26 cities outside of Utah and 23 Utah cities" https://www.fairvote.org/rcv#where_is_ranked_choice_voting_u...

Approval in the US: Fargo and St. Louis

RCV campaigners hijacked public perception of the science. As an analogy, people don't believe in global warming because they have evaluated the evidence; they believe it because the scientists tell them to. RCV campaigners skipped the scientists and manufactured a false consensus around RCV. So RCV is what the common person associates with voting method reform.

Sure, and I can't criticize the model with any certainty, or make my own firm claims, without doing modelling myself. But looking for concerns is fine. I can't say the model is wrong in this aspect even though it looks odd. However, I can think of alternative models which would definitely have the opposite result. (For example, squared distance as utility.)

>Of course there is. The ideal would be a model that is exactly like real life.

Practical considerations make this impossible. Validation of a model by comparing to reality only works for small subsets of voting theory at the moment, and weakly so. A complete model like this one has no hope. And if comparison to reality is the goal, I can already see gaps where it diverges from the best-known theory, by omission of features. So taste is important. I have my own preferences about what features are important, which others will disagree with. These preferences will be non-falsifiable for at least 50 years while we wait for experiments to catch up.

For example, I assume you mean affine scaling of utility into the [0, 1] range when you say "Normalization of sincere utilities". But we might add to each election a candidate whose goal is to nuke the country itself. His utility is massively negative for every voter. However, it isn't to my taste for this candidate's presence to distort "honest" votes; I'd rather there be some clamping at the ends. So even this "obvious" choice is not canonical.

>Be specific. It just sounds like you don't understand how it was done.

Yes, that's the problem. I shouldn't have to go through code to understand necessary things. An expert in the field should be able to follow along by reading the paper. Although, this is not a paper, so holding it up to that standard is unfair. But the gaps in what I can figure out make the results non-interpretable in some of the aspects I most care about.

I think one of the results you most care about, from the model, is that Approval measures better than ranked choice. I think that's still a fair conclusion to arrive at.

A typical example where bullet voting should occur is a down-ballot election. Examples of low-engagement elections at https://en.wikipedia.org/wiki/Approval_voting#Other_organiza... bear out that these elections degrade in behavior to plurality. However, it can't be determined just from this behavior that IRV or Condorcet would do any better.

> https://www.rangevoting.org/BulletBugaboo.html

1-3 have major engagement from voters and I do not expect approval voting to suffer unduly from bullet voting. 4, an alumni association, should. However, the question then becomes, what about other voting methods? Because if the voter has weak information on the other candidates, he may simply bullet vote under all voting methods. In this context, Score needs testing. Voting values between 0 and 1 would give a much better idea of how voter engagement is affecting their votes, rather than their second choice always collapsing to zero. This is why I claim that it is "unclear to what extent it affects score methods in practice". Granular cardinal scores in even one such election would mean a lot for determining voter behavior.

I do not support FairVote's arguments and consider them dishonest. Notice that in my original post, I carefully describe bullet voting as a consequence of low engagement, which is supported by past real-world low-engagement elections. This is very distinct from FairVote's arguments, which are not sophisticated and demonstrate a willingness to make blind claims. The distortions that FairVote argues for require more justification than they give.

> https://electionscience.github.io/vse-sim/VSEbasic/

I don't believe much in that simulation. While I agree with your characterization of it as the "best computer modeling" in voting theory, it would still be considered a fatally flawed paper under the standards of most fields, and I faced some heavy obstacles when trying to analyze it. I think this field needs more/better academics. I described some of my objections at https://news.ycombinator.com/item?id=27612876

I don't know what is up with 1-sided score vs 1-sided approval. That is an interesting observation you made and does look suspicious; we should expect the distortion of power to be greater than the loss of honest expressiveness.

When I looked at this project a year ago, I found other strange behaviors that led me to ignore its results, similar to how you are comparing its graphs to your intuition. There is no "correct" model in voting theory. Models are neither good or bad, only to the author's taste or not. For example, with voters on a single dimension, should utility of a candidate be the squared distance, or the absolute distance? How should a candidate's positions be distributed, is n-dimensional Euclidean space good? Should some candidates be straight-up better than others, and what should the distribution be? How should voters be clustered, if at all? What does "strategic" voting mean in each context? What does "honest" voting mean in cardinal contexts?

The author's discussion of methodology was lacking, and the missing priors make its results partially meaningless. I also had no success looking through the code for even basic behavior, because it was too hard to navigate. However, this is only a graduate project, and most graduate projects (including mine) are of lower quality than this one, so I wasn't upset.

In score systems, voters can sometimes be made honest to an extent. For example, on a book rating site, you probably don't rate all books 0 or 5; you use the full scale. Your ratings have greater self-value than your desire to tip the global score. While your assertion is reasonable that it is not the expected "honest" voting behavior, it is also a behavior worth looking at. This goes back to discussion of methodology; all these assumptions need to be documented, rather than buried in hard-to-read code.

I think STAR is slightly worse than Score, is comparable to Approval, is much better than IRV, and is likely better than Condorcet methods. STAR has some odd behavior which can be explored in a 3-person race. Score has less-problematic behavior caused by risk-taking with equilibrium voters (not like voters behave in any way similar to Nash equilibria though, and who knows what the real-world behavior will be).

However, STAR's main benefit may be in overcoming political resistance, if its properties are simpler to convince voters. Majority criterion sounds nice even when it is inefficient. Much like Top Trading Cycles losing to Gale-Shapley in school choice algorithms (excluding Boston).

A more concrete example is that anthropology was introduced to statistics in the 1960s. [2] Before that, many papers use storytelling as their main evidence. They reflect the thesis the author wants to present, and experiments were not available to anthropologists as a tool. This makes it hard to tell which papers should be taken seriously and which should not. You either trust the author or don't; you can't analyze his procedure. So papers cannot build on each other, and progress is halting. The quality of anthropology papers rises substantially after 1970, thanks to experimental methods. Scientific imperialism has turned modern anthropology into one of the paragons of social science, with enormously interesting results today.

Here's a fun experience in which you can discover this phenomenon for yourself. Synthetic reverbs rely on delay lines, and the question is, how to choose the delay lengths? Every reverb engineer has spent impressive amounts of time manually tuning their delay lengths in random ways, then listening to the result, then tuning some more. The task is to choose a set of N real numbers such that their lattice points (linear combinations with nonnegative integral coefficients) are maximally separated in the best sense you can. For example, if N=2, then 1 and 5/7 would be unsuitable, because 1 x 5 = 5/7 x 7 = 5, and exact overlaps are poorly separated. Irrational numbers do better, because they at least never have exact matches. Manual tuning is hard because tweaking one number to erase an overlap will introduce a different overlap.

You, the non-mathematical reader, can now solve this problem by navigating to https://en.wikipedia.org/wiki/Markov_constant, reading the definition of "quadratic irrationality", and picking square roots of squarefree integers as your solution. Congratulations, in 5 minutes, you solved most of a major open problem, on which reverb engineers have spent 10000 hours and made no progress. First person to write a paper gets credit. [3]

There's a general malaise in reverb engineers regarding mathematics and science. Some people already have the background information to solve the problem, as they know 1.618 is the solution for N=2 (which leads to Hurwitz's theorem, then Markov constants). But then they advocate for 1.618^x for N>=3, which means they are doing cargo cult mysticism around magical numbers, rather than actual mathematics. Note that powers of 1.618 are linear combinations of smaller powers, so their answer is the worst possible.

[1] I'm just joking. But only partially joking. Hardy-Weinberg principle. [2] https://www.annualreviews.org/doi/abs/10.1146/annurev.an.14.... [3] I'm joking for real this time. Writing a paper will do nothing.

The responsible vulnerability was in Java applets, which I had disabled for security reasons. But Java secretly re-enabled itself after updating. I kept Java uninstalled for a long time after that.

On the other hand, trailer homes also centralized reputation into a few major companies, and nevertheless had severe issues with VOCs and formaldehyde. These trailers were unlivable, so reputation mechanisms didn't manage to raise quality.

There's no underground foundation. I was shocked when they stacked buildings, although there are no stairs to get to the second floor. It's questionable whether a good foundation also serves as a good second story floor (weight? sound? flexure?) but I guess being modular can be worth it. It looks like a saner version of the shipping container home, which is popular in imagination and unpopular in reality.

The attempted niche of this particular company is its price, which is half of comparable prefab homes. It's very low-end. The windows have no shades and no meshes. Sunshine will glare during the day, with no privacy. Open plan, kitchen in living room. Roof seems at risk of leaking. Unclear what the material quality is like (VOCs, heat and sound insulation, appliances). They haven't shown any of the hookups (water, electrical, sewer, internet). No gas. Isn't yet available for purchase.

Here's a link speculating on the impact of Apple's IDFA deprecation on Facebook, from January: https://mobiledevmemo.com/facebook-may-take-revenue-hit-from... It carries strong parallels to what is in-process at Google.

A few ad experts I follow say that Facebook and Google will claim increasing dominance in the ad space. Their logic is that as each ad vendor loses third party information, comparative advantage will favor companies with strong first party information. This will drive more ads toward Facebook as a relative share, even if they are still losing overall. I have no information to judge to what extent this claim holds.

At my graduate institution, the undergraduate attitude was quite different; socializing was important, learning the minimal amount was efficient, and cheating was socially accepted (by peers, not necessarily the profs).

I was shocked at the difference in academic achievement this caused. The tests at the second institution were far easier, yet people were failing them. This was a difference between scoring 80% on nontrivial proofs, vs scoring 60% on one-step theorem applications. I don't think the incoming freshmen were vastly inferior. Rather, the culture at the first institution forced students to apply themselves through peer pressure, because that's what they perceived everyone else to be doing, and also what was valued. My advisor at the second institution seemed quite pessimistic about changing its culture, and viewed the first institution as a rare exception that would not be emulatable. He did agree on the extent of the difference.

From the prof perspective, this is all irrelevant because it's quite rare for a school to develop a good culture, so it's not the experienced reality for most profs. Asking to become faculty at a top institution where the undergraduates also all care about academics is like wishing for unicorns.

* Some selection bias here: I took grad classes, so these were the only math majors I had exposure to. Although, these were the only classes available because the school was small, so I'm not sure what an undergraduate could take if he didn't want to take grad classes.

It really opened my eyes that google was willing to hold my emails hostage. Google's roadblock was unrelated to security, because an attacker could provide a new phone number too.