“The needs of the many,” privilege, and power

There’s a certain set of sentiments which undergird a lot of thinking in engineering, and especially engineering about data. You want a method which has good performance “on average” over the population. The other extreme is worst-case, but there are things you can only do in the average case. By focusing on average-case gain, you get a kind of “the needs of the many outweigh the needs of the few” way of thinking about the world.

The needs of the many outweigh the needs of the one... or the few?

The needs of the many outweigh the needs of the one… or the few?

Now in the abstract land of mathematical models and algorithms, this might seem like a reasonable principle — if you have to cram everything into a single population utility function you might as well then optimize that. However, this gets messier when you start implementing it in the real world (unless of course you’re an economist of a certain stripe). The needs of the many are often the needs of the more powerful or dominant groups in society. The needs of the few are perhaps those who have been historically marginalized or victimized. Extolling the benefits to the many is often taking a stand for the powerful against the weak. It’s at best deeply insensitive.

Two instances of this have appeared on the blogosphere recently. Scott Aaronson blogged recently about MIT’s decision to take down Walter Lewin’s online videos after Lewin was found to have sexually harassed students in connection with the course. Scott believes that depriving students of Lewin’s materials is a terrible outcome, even (possibly) if he were a murderer. Ignoring the real hurt and trauma felt by those who are affected by Lewin’s actions is an exercise in privilege — because he is not hurt by it, he values the “the good of the many” trumping the “good of the few.”

The whole downplaying of sexual harassment as being somehow “not serious” enough to warrant a serious response (or that the response “makes the most dramatic possible statement about a broader social issue”) in fact trivializes the whole experience of sexual violence. Indeed, by this line of argument, because the content created by Lewin is so valuable, it may be ok to keep online even “had [he] gone on a murder spree.” The subtext of this is “as opposed to merely harassed some women.” I recommend reading Priya Phadnis on this case — she comes to a very different conclusion, namely that special pedestal that we put Walter Lewin on is itself the problem. Being able to downplay the female victims’ claims is exercising the sort of privilege that members of the male professoriat (myself included) indulge in overtly, covertly, and inadvertently. If STEM has a gender problem, it’s in a large part because we do not pay attention to the ways in which our words and actions reinforce existing tropes.

The second post was by Lance Fornow on dying languages in response to an op-ed by John McWhorter on why we should care about language diversity. Lance thinks that speaking a common language is a good thing:

I understand the desire of linguists and social scientists to want to keep these languages active, but to do so may make it harder for them to take advantage of our networked society. Linguists should study languages but they shouldn’t interfere with the natural progression. Every time a language dies, the world gets more connected and that’s not a bad thing.

I guess those poor bleeding-heart social scientists don’t understand that those languages are dying for a good reason. The good of the many — everyone speaking English, the dominant language — outweighs the good of the few. This attitude again speaks from a place of privilege and power, and it reinforces a kind cultural superiority (although I am sure Lance doesn’t think of it that way). Indeed, in many parts of the world, there is and continues to be “a strong reason to learn multiple languages.” By casually (and incorrectly) dismissing the importance of linguistic diversity, such a statement reinforces a chauvinist view of the relationship between language and technology.

We start with desirable outcomes: free quality educational materials that lower the barrier to access or speaking a common language to help facilitate communication and cooperation. By choosing to focus on those outcomes and their benefits to the many, we value their well-being and delegitimize the harm done to others. If we furthermore are speaking from a position of power, our privilege reinforces stigmas, casting a value judgement on the rights, experiences, and beliefs of the few. It’s something to be careful about.

Linkage

I will post more about Allerton soon (I’m still on the road), but I wanted to clear out some old links before doing that. I’m starting my new gig at TTIC this week, and the last few weeks have been a whirlwind of travel and internetlessness, so blogging has been curtailed.

And a (not-so-recent) tour around the ArXiV — I haven’t had a chance to read these yet, but maybe once I am settled…

Shannon theory helps decipher Pictish?

Well, if not decipher, at least claim that there is something to read. A recent paper claims that Pictish inscriptions are a form of written language:

Lo and behold, the Shannon entropy of Pictish inscriptions turned out to be what one would expect from a written language, and not from other symbolic representations such as heraldry.

The full paper has more details. From reading the popular account I thought it was just a simple hypothesis test using the empirical entropy as a test statistic and “heraldry” as the null hypothesis, but it is a little more complicated than that.

After identifying the set of symbols in Pictish inscriptions, the question is how related adjacent symbols are to each other. That is, can the symbols be read sequentially? What they do is renormalize Shannon’s F_2 statistic (from the paper “Prediction and entropy of printed English”), which is essentially the empirical conditional entropy of the current symbol conditioned on the past symbols. They compute:

U_r = F_2 / \log\left( \frac{N_d}{N_u} \right)

where N_d and N_u are the number of di-grams and un-grams, respectively. Why normalize? The statistic F_2 by itself does not discriminate well between semasiographic (symbolic systems like heraldry) and lexigraphic (e.g. alphabets or syllabaries) systems.

Another feature which the authors think is important is the number of digrams which are repeated in the text. If S_d is the number of digrams appearing once and T_d is the total number of digrams, they use a “di-gram repetition factor”

C_r = \frac{N_d}{N_u} + a \cdot \frac{S_d}{T_d}

where the tradeoff factor a is chosen via cross-validation on known corpora.

They then propose a two-step decision process. First they compare C_r to a threshold — if it is small then they deem the system to be more “heraldic”. If C_r is large then then do a three-way decision based on U_r. If U_r is small then the text corresponds to letters, if larger, syllables, and larger still, words.

In this paper “entropy” is being used here as some statistic with discriminatory value. It is not clear a priori that human writing systems should display empirical entropies with certain values, but since it works well on other known corpora, it seems like reasonable evidence. I think the authors are relatively careful about this, which is nice, since popular news might make one think that purported alien transmissions could easily fall to a similar analysis. Maybe that’s how Jeff Goldblum mnanaged to get his Mac to reprogram the alien ship in Independence Day

Update: I forgot to link to a few related things. The statistics in this paper are a little more convincing than the work on the Indus script (see Cosma’s lengthy analysis. In particular, they do a little better job of justifying their statistic as discriminating in known corpora. Pictish would seem to be woefully undersampled, so it is important to justify the statistic as discriminatory for small data sets.

Romanian diacritics

I came across this blog post today while trying to figure out how to write the Romanian breve (the symbol ă) in a document, and it was an amusingly angry rant about Romanian orthography. The fact that the Romanian currency even got it wrong is pretty funny. But it seems a bit like a futile battle; things always change and I bet the orthography gets merged eventually. I, for one, miss the ess-zett (ß) in German, but it’s gone the way of the dinosaurs.

That would be a great name for an diacritic mark — a dinosaur. A stegosaurus sitting on top of a U. But how would it be pronounced?

You say Tschebyscheff, I say Chebyshev

As I reread the Burnashev-Zingagirov paper on interval estimation today, I came across a new (to me) spelling of the mathematician Chebyshev‘s name. I found a page with variant spellings, including

  • Chebyshev
  • Chebyshov
  • Chebishev
  • Chebysheff
  • Tschebischeff
  • Tschebyshev
  • Tschebyscheff
  • Tschebyschef
  • Tschebyschew

I know that “Tsch” comes from French/German transliterations. But today I saw “Chebysgev,” which is a totally new one to me. Where does the “g” come in? The name is actually Чебышев, which may or may not show up depending on your Unicode support.

UPDATE : Hari found “Tchebichev” in Loève’s Probability Theory book.

free speech and America-centrism

Chris Bertram over at has a post on speech regulation with which I’m not sure I agree, but I do wholeheartedly agree with this sentiment:

The Americans have a long tradition of trying to discuss these things using the language of an 18th-century document. Given the difficulties of shoehorning a lot of real-world problems into that frame, that gives them a long history of acrobatic hermeneutics somewhere in the vague area of free speech. Some of it is even relevant. The trouble is that many Americans (at least the ones who comment on blogs!) can’t tell the difference between discussing the free speech and discussing the application of their constitution.

Not only true on blogs, but in person as well.