Thursday, October 19, 2017

Weapons of Math Destruction

Version 0.4

Review of and Reactions to Weapons of Mass Destruction by Cathy O'Neil

Quite a good book, though I have a number of nits to pick. Overall the book is worth reading and I'm glad to give it the same strong recommendation that led me to read it, but there are places where her focus seems too shortsighted, especially in terms of the underlying motivations of the problems.

The largest nit is probably the title, not so much directly, but as used in abbreviated form within the text. The title is actually catchy and provocative, though I think it falls short of witty. However, in the text it is reduced to WMD, and that's a serious problem because there is a competing and quite well established definition of WMD. Weapons of Mass Destruction are still a huge problem and frequently mentioned in the news and books. There are good reasons we read about WMDs, but that means that each time WMD appears in this book the reader has to stop and remember the crucial distinction here. Actually, it would be even worse if she had focused more on the mass destruction that can be caused by her WMDs. She didn't go that far in spite of her focus on the scalability of the the weapons. (There's a secondary problem that she probably couldn't have considered, but for Japanese readers in particular, the spoken words "math" and "mass" are hard to distinguish because Japanese does not use the "th" sound. In other words, to a Japanese ear, both "math" and "mass" sound like マス。)

The reading would have been much easier without the conflicting WMD references in the test. Perhaps WM for Weaponized Math or MWs for Math Weapons? The kind of math is actually limited to statistics, so another option could have been SWs for Statistical Weapons. Lots of possibilities beyond the catchy but conflicted WMDs. Just hard for me to recommend this book without a warning about being a more difficult read than it needs to be.

Now for my page-linked comments (though I resisted noting pages until I was halfway through):

On page  115 I was provoked by the passage "a quality computer, like an IBM Selectric", since that was a typewriter, not any type of computer. That was the reaction that went too far...

On page 130 she is talking about a blurred focus on "efficiency and profitability" without justifying the inclusion of efficiency. This reminded me of corporate cancerism, though without the clarify of the singular focus on profit that modern corporations claim they are legally required to pursue.

Page 134 starts considering the misguided destruction of public education in 1983, but I was mostly reminded of "Figures don't lie, but liars figure" and the book Lies, Damn Lies, and Statistics from 1976. I felt she wasn't seeing the real problem because she wasn't considering the motivations. She is focused on the attacks on unionized teachers, but she didn't even consider why they would be targeted. I would actually agree that they were part of the motivation was that progressive teachers can encourage progress and change, and conservatives are fundamentally opposed to that, but I think she failed to see the larger attack in which public schools were divided and conquered. The result was a tiny division of good schools to keep up the hopes of the good teachers and the truly concerned parents, while the lion's share of public schools became obedience schools you wouldn't send your dog to. The underlying motive for this wholesale destruction was mostly to reduce property taxes, though the rich landowners were glad to use and exploit such public-school-haters as religious fanatics who had other reasons for wanting the public schools to be destroyed. Though she talks about problems with voting and democracy at various places in the book, she apparently doesn't see how the mushrooms were cultivated in those obedience schools so that Putin (with Trump's help) could harvest them as mushroom voters in 2016.

Page 136 on Simpson's Paradox was quite enlightening, but mostly a reminder of how liars figure. It doesn't matter what the data shows when there are external motivations driving the analysis and the desired outcomes have been predetermined. The liars will find some way to slice and dice it and they are probably planning to take the money and run before the suckers figure it out.

Page 143 was about selective treatment in call centers, and made me paranoid about how I get treated these days. Also reminded me of the extreme form practiced at Apple, where I believe they now block negative comments from even being posted in their discussion forums.

On page 146 she emphasized that it was wrong to study "How have people like you behaved in the past?" She said the ideal question would be "How have you behaved in the past?" Actually, the ideal question would be "How are you going to behave in the FUTURE?"

Page 162 is discussing insurance, but I felt there was general confusion. The basic problem is that it only makes mathematical sense to insure against unlikely events, not likely events. Trying to insure against inevitable events is a kind of oxymoron, because there can't be any profit there if all of the policy holders are guaranteed to collect at some point. They might as well save their money and cut out the insurance company overhead.

Page 180 is talking about viral encouragement of voting via Facebook, but mostly it sounds ridiculously naive in light of the 2016 election. There was no mention of Twitter in the book, either. Page 185 sounded even more naive when she says that she has "no evidence that the companies [Facebook and Google] are using their networks to cause harm." Page 186 sounds almost ridiculous when you consider how Trump just ignores all of the conflicts and counts on each of his contradictory tribes to believe he is telling that particular tribe the truth while lying to all of the others.

On page 205 she is talking about an "oath" that data scientists should take, but the notion seems completely implausible because of her own emphasis on scalability of the WMDs. If a WMD is scalable, then all it would take is one violator of the oath to render it useless. This reminded me again of the moral neutrality of the technology (though she didn't use that phraseology) while the people are going to use any technology for good or bad purposes.

Thursday, September 21, 2017

山梨県のひみつ

Not an actual book review, but just a reactionary letter sent to the publishers of a book that especially annoyed me:

Again, my apologies for writing in English. My Japanese writing is poor, notwithstanding having read over 100 of your ひみつ books.

I just finished 山梨県のひみつ and I feel I have to say something. More so since I actually lived in Yamanashi many years ago.

The smallest problem is one that I've called to your attention before: Unreadable fonts. This is most often a problem with the まめちしき, especially the フリガナ where the little 〃 and ゜ can be hard to tell apart. However, this book had MANY pages of small print that made the problem much worse. I actually wound up running those pages through a high resolution scanner and displaying them full-screen on my 28-cm monitor, and even then they were barely readable in places.

That leads to the middling problem, which was the book took MUCH more effort than any previous volume of ANY of your series, which I've been reading for many years. In general, the ひみつ books are a pleasure to read, but this one was NOT. If you don't believe me (or if you don't care), then I strongly urge you to ask ANY of the children (your intended audience) who have actually read this volume.

Sometimes less is more, and in this volume there was LOTS of stuff that could have been cut, which also would have allowed you to use larger fonts for the remaining material while making the book more pleasant to read. For example, all of the America stuff was unimportant and distracting and should have been cut. The details hours and fees of the museums are not needed here, but if you insist, all of that stuff could have been reduced to a one-page table at the end of the book. Lots of unneeded redundancy, especially around the high-speed train.

Lastly, the big problem: I felt personally offended by the use of katakana for the Japanese words of the two American children. Hard to read and it feels downright racist, too. Perhaps the biggest difference between America and Japan is that Americans think that anyone can learn to speak English, while the Japanese think no one can learn to speak Japanese unless they are born Japanese.

Thursday, February 23, 2017

Reactions to General will 2.0: Rousseau, Freud, Google

Reactions to General will 2.0: Rousseau, Freud, Google by Hiroki Azuma (一般意志2.0 ルソー、フロイト、グーグル)


In summary, a tremendously provocative and rather interesting book, but too flawed and immature to be important or influential. It makes me feel like I have to start with rationalizations about the book's limitations. I think there're three kinds of problems that affected this book.

One is that the translation itself seems somewhat suspicious, Either because the material is so difficult or possibly because the translators own philosophies were brought into the picture, especially if either of them is a libertarian. In my page-based notes I will include a few specific questions about the translation, but I think there's more a general problem manifested in the confusion of proper nouns with generic usages. Quite frequently the book uses terms such as "general will 2.0" that should be capitalized and treated as proper nouns, but which appear in lowercase. It is barely possible that this reflects a feeling that the frequent capitalization would have made the book feel too heavy in some way, but I think it much more likely that it reflects the lack of capitalization in the original Japanese. (There are no capital letters in Japanese.)

The second problem could be described as a kind of trauma due to the disastrous earthquake and tsunami of 2011, immediately followed by the nuclear-power-plant disaster in Fukushima. My own trauma was relatively minor, just that I voided elevators for a few months, but perhaps the author had tighter connections to Fukushima or for some other reason felt the disaster more deeply. He gives the impression in the introduction that he was forced by the earthquake into rethinking many of the points raised in the book in ways that prevented him from merely polishing the original essays. His choices were to present them almost intact from the magazine serialization or completely rewrite the book.

The third problem can be described as a kind of forced extension beyond the original conception, especially as regards Rousseau. He is trying to rework the philosophy of Rousseau in a way that goes beyond the plausible implications of the original philosophy of the "social contract". I think it started as an interesting thought experiment, but he went so far that he at some point even the author understood that the connection to the original was overstretched. Even with the stretching, it's thought-provoking, but I felt it was not a mature conceptualization. There was some additional confusion caused by the layering in of later philosophers and thinkers, especially as regards Freud.

So here are my page-based comments:

On page vii I saw the first thing that made me worry about the translation. Near the bottom he says "I was able to assume the place ..." in a way that sounds quite pompous and assertive, even aggressive. I suspect the original Japanese may have represented a gentler conception that should've been translated into something like "I found myself thrust into a place ..."

On page xii he notes that it is September 11, 2011, which is exactly half a year since the disaster of the great earthquake. As an American I was struck but that the date was also exactly 10 years after the great and similarly traumatic disaster of 9/11 in America. It would have been even more striking as the double anniversary. (The 9/11 disaster was mentioned on page 60.)

On page 6 he talks about the Google's mission statement of organizing the worlds information. My main reaction was negative in light of my reconsideration of that mission statement in favor of giving priority to accessing the advertising information with the new utility metric of the paying advertisers' profits. I wonder if his generous and friendly interpretation was still plausible when this book was written?

At the bottom of page 11 was another sentence that made me wonder about the quality of the translation. It uses a very unusual word, "sublated", in such a way that it could not be clearly interpreted from the context. Perhaps a dictionary-based translation? Following the literal Japanese, but the result was not clear in the actual context. Better writing or a better translation should have made this concept clear. There are other examples in the book where free translation was clearly used. For example the "two heads are better than one" expression is used instead of a direct translation of the Japanese expression about the three men being wiser then Buddha. (三人寄れば文殊の知恵 was referenced in a note.)

On page 53 there is a confusing note about the two Japanese forms of 意思 and 意志, which are homonyms with closely related meanings. (One kind of 同音異義。) However my main reaction to this footnote was to realize that all of the Japanese intrusions in the book would have benefited from the inclusion of proper Japanese. The use of Romaji was as usual ambiguous and confusing. If you don't already know the Japanese, then the Romaji is just gibberish, and if you do know the Japanese, then the Romaji is just intrusive.

On page 69 he's discussing the shallowness of conversations on the infamous Nichanneru (2ちゃんねる?). Reminds me again of Nicholas Carr's excellent book The Shallows.

On page 70 the second note is talking about the creator of the Google's Japanese input system, someone named Taku Kudo. One of my reactions was to wonder if he was might be related to the security expert Kudo-san at IBM Japan, but mostly I was curious about the extra features of Google Japanese input.

What struck me about page 73 was his essential confusion about the nature of freedom. However I think it helped to trigger me to think more about the meaning of constraints based on reality within the context of my own equation defining the most important sense of freedom. [#1 Freedom = (Meaningful - Coerced) Choice{5} ≠ (Beer^4 | Speech | Trade) in my favorite sig.] Not sure if this is when I started wondering about some way to work the word constraint or reality into my sig, but it also reassured me about the usage of "Coerced" there.

My main reaction to page 79 was probably do to his apparent confusion about the you effective use of personal time. This is also another intersection with The Shallows.

Page 81 i talking about island universes from a libertarian perspective, but this is a concept I now map to do-it-yourself brainwashing.

Page 86 represents another other intrusion of libertarian fantasies. He's trying to rationalize minimizing government without considering the essential blindness of masses of people. The kinds of data he is collecting here are fundamentally incapable of revealing what should be done, incapable of dealing with the notion of change itself, but only capture the static conditions. I7m doubtful that the author understands that leadership requires unified vision, though this might be an overlaying of the libertarian conceptions. There's also some confusion with his use of the word "database" throughout the book, where he was probably thinking about something like "big data". Perhaps the use of big data came later, so he was forced to use the approximating word database?

I had two reactions on page 96. The main one was feeling his interpretation of Google PageRank was quite shallow and even inaccurate. I also had trouble with his interpretation of Freud here.

Not certain what caught my attention page 113, but perhaps that is where I realized the confusion between database and big data? The closing paragraph of that chapter was quite confusing to me, though that might be another translation problem.

Page 115 he had me thinking about reality-based constraints in relation to freedom.

On pages 136 and 137 he made me feel like he was confused about how the wisdom of crowds works, even though he refers to that book. The independent perspective of the individuals within the crowd is key to avoiding mob-based decisions.

The third note on page 139 made me think of time-based economics, although that general topic is clearly beyond his thinking in this book. I think that was the main problem that caught my attention on page 145, too.

On page 150 he's talking about the Japanese website Niconico, which has a real-time chat mode similar to YouTube's. He's talking about the problem of an overabundance of comments on the right side. This make me think about an obvious solution. Most of the comments should never be displayed, though perhaps they could be saved somewhere else for later reference. The displayed comments should be throttled to a slow speed to make sure they can be read. After each comment that makes it onto the display there would be some number and buttons. The first number would represent how many comments had been skipped before accepting the displayed comment. This would give a real-time indicator of the activity of the discussion. A thumbs-up and thumbs-down button would let the audience try to push the comment up or down the list, with a number for the net value. If the comment is accumulating a lot of positive reactions, then will tend to rise in the stack of displayed comments, but if it's getting negative reactions it will sink. There should also be a button for "the subject has changed". If enough viewers click on that one, then the comment will disappear to make room for a new comment. This system would then semi-automatically create a time-based list of the most important comments associated with each part of the video. More ideas and details available upon request, but that offer feels like a joke these days...

The discussion of Twitter on pages 182 to 184 was also thought provoking. It made me think of a new way to make Twitter much more interesting and useful. You should be able to sort and group the accounts you are following so that the different kinds of information appear in separate lists. On a large screen computer you would be able to display the lists side-by-side. For example the Twitter feed of your family members and close friends might be in the first column on the left side, while the second column might be for news sources and the third column for celebrities you are interested in. On a small screen device a sideways flick could switch between the columns representing your groupings. Another new idea of the sort I like to think about...

On page 204 he suggests that Google and Apple are influenced buy their shareholders. Seems to me to be a remarkably naïve statement, based on my firsthand experience as an Apple shareholder and most starkly in a recent interaction with Sony. They are not interested in new ideas or any sort of "guidance" from shareholders. That's part of a more general problem and made me think his economic models are too naïve by half.

Those are my page-based reactions to the book. They sound somewhat critical, but I'm still interested in many of the topics he raised. Not sure if I can actually recommend the book, but I'm thinking of visiting the coffee shop (or maybe it's more of a discussion salon) that the author manages.