Congratulations to Justin and Jon

Justin Thaler and Jon Ullman had back-to-back thesis defenses today.  Both talks were excellent -- Justin's on his work on verification methods (for cloud computing), and Jon on his work on differential privacy.  While there is still some small paperwork matters (like the actual theses to turn in), I think we should start calling them Doctor Thaler and and Doctor Ullman, just so they start getting used to it.

Congratulations Justin and Jon!

Calling Out Stupidity

While I greatly enjoy working at Harvard, there are many painfully stupid things said and done by people here -- as I suppose is the case anywhere -- that either I don't feel merit commenting on or I don't feel it's appropriate to comment on.  (And, of course, I'm sure that sometime in the past I've done or said things that others find stupid, and I'm very happy they aren't blogged about.)

However, the recent comments by Niall Ferguson are out in the public, and so over-the-top stupid that even he quickly realized how stupid they were.  And I think it's important to point out, prominently, how stupid they are, because the idea that either childless people or gay people somehow have a discounted view of the importance of the future deeply offends me, and somehow the idea that someone prominent in my workplace would say (or believe, or say in stupidity without really believing) such a thing has made my week substantially sadder.  And so I feel the need to point it out, ideally not to sadden anyone reading this, but to emphasize how sad and stupid the comments were.   

A Boy and His Atom

Some researchers at IBM are so good at playing with atoms, they decided to make a movie (called A Boy and His Atom), by moving atoms.  Cool stuff.  Computer science connections:  implications for storage.  Personal connection:  the spouse of the scientist leading the group who made the video is a friend from high school. 

Some Recent Books

If you've been reading the other computer science blogs, you know that there are some new books out.

First is the The Golden Ticket: P, NP, and the Search for the Impossible, by Lance Fortnow, about the P vs. NP problem.  I got sent a review copy;  I haven't read it yet, but I'm testing it out by having my oldest daughter read it and tell me what she thinks, and I'll read it along with her.

Then there's Quantum Computing since Democritus, which covers complexity theory, quantum computing, cryptography, and a bunch of other stuff.

But there's more!  Thomas Cormen, of the famous CLRS Introduction to Algorithms book, has a new book out called Algorithms Unlocked, which seems to be a friendly introduction to the basics of algorithms.


On the popular books front, Eric Schmidt (yes that one) and Jared Cohen have a book out on The New Digital Age: Reshaping the Future of People, Nations and Business.

Also, the book Nine Algorithms That Changed the Future: The Ingenious Ideas That Drive Today's Computers, from some time ago, is out in paperback.



On the textbook side, last year Steven Gortler of Harvard introduced a new textbook for computer graphics that looks really good.  And of course, there's always that Mitzenmacher/Upfal Probability and Computing: Randomized Algorithms and Probabilistic Analysis to buy if it's not already on your shelf...

Thanks for the Memories....

I'm off to a meeting next week, so today was my last day of class for Computer Science 124 (Algorithms and Data Structures) for the semester.  For two reasons, today feels important. 

First, because I'll be on sabbatical next year, I won't have to teach another lecture for what seems to be something like 18 months.  I haven't had that much time off from teaching since I've started.  I'm looking forward to it, especially as this year the administrative job has slowed research.  I'm hopeful the sabbatical time can be put to use productively on the research side.

Second, more nostalgically, I've taught this class for the last 15 years.  To me that feels like a record of some sort.  When I come back from sabbatical, the plan is that someone else will be teaching the course, and I'll have to find something new and entertaining (for me, and maybe for the students) to teach.  I admit, my preference would be to teach this class pretty much forever.  After 15 years, I finally think I'm starting to understand all the material and how it fits together.  I'll miss the class.  While my head knows that it is probably good for both me and the class to separate, my heart hurts a little, giving it up and moving on.     

Given the excitement of once again being done with teaching for the academic year, I'm sure I'll get by.  (As I explain to the students, the only ones happier than they are that classes are ending are the faculty...)  Things change -- obviously, for me, sometimes very slowly -- and I'm optimistic that whatever I decide to teach next, I can make it a course I'll enjoy teaching for say the next 15 years.

Guest Post by Mark Bun and Justin Thaler

Michael asked me and Mark to write a post on our paper "Dual Lower Bounds for Approximate Degree and Markov Bernstein Inequalities", and we're happy to do so. As the title suggests, our paper focuses on understanding a certain measure of the complexity of a Boolean function, known as approximate degree. Given an n-variate Boolean function f, the approximate degree of f is the least degree of a real polynomial that pointwise approximates f to accuracy 1/3 (the choice of the constant 1/3 is by convention -- the particular constant is inconsequential).

Aside from being a natural complexity measure, approximate degree has a surprisingly wide range of applications throughout theoretical computer science. A very partial list includes:

1) Lower bounds on approximate degree yield lower bounds on quantum query complexity.

2) Lower bounds on approximate degree have recently been used to resolve long-standing open questions in communication complexity (see here for a survey from 2008, and some more recent works here and here).

3) Upper bounds on approximate degree underly the best known agnostic learning algorithms.

Despite the range and importance of these applications, large gaps remain in our understanding of approximate degree. The approximate degree of any *symmetric* Boolean function has been understood since Paturi's 1992 paper, but once we move beyond symmetric functions, few general results are known. One function that has served as a symbol of our lack of understanding is the two-level AND-OR tree, the function computed by a CNF (i.e. an AND of ORs) with all gates having fan-in sqrt(n). Over the last couple of decades, there has been a line of work proving incrementally larger lower bounds on the approximate degree of the AND-OR tree, with the best previous lower bound being Omega(n^{3/8}) due to Sherstov. Our main result in the paper resolves the question completely by establishing an Omega(sqrt(n)) lower bound, matching an O(sqrt(n)) upper bound due to H\oyer, Mosca, and de Wolf. (Sherstov recently independently obtained the same Omega(sqrt(n)) lower bound using related techniques). Our second result is to give a new proof of Paturi's lower bound for symmetric Boolean functions.

Let us say a bit about how our proofs of these two results work. Traditionally, approximate degree lower bounds have been proven by a technique known as symmetrization, which transforms questions about multivariate polynomials into well-understood univariate ones. In Step 1, one supposes there is a n-variate polynomial p that pointwise approximates f to error 1/3, and turns it into a univariate polynomial q in such a way that deg(p) >= deg(q). In Step 2, one shows that q must have high degree, which means p must have high degree as well. Unfortunately, the symmetrization step of moving from p to q often 'throws away' information about p (after all, p is an {n choose d}-dimensional object, and q is only a d-dimensional object), so one sometimes runs into problems in completing Step 2.

Recently, there has been progress in moving beyond 'lossy' lower bound techniques like symmetrization. Our paper uses the following approach, which is completely lossless. You can straightforwardly formulate the approximate degree of a function f as a linear program: the approximate degree of f is at least d if the value of the following LP is larger than 1/3:

min \epsilon

s.t. |f(x) - \sum_{|S| <= d} c_S chi_S(x)| <= epsilon, for all x in {-1, 1}^n

c_S in \mathbb{R}

Here, chi_S(x) is the parity function over variables in the set S. This program is just asking you to construct a degree d polynomial that minimizes the distance to f under the L_{\infty} norm.

If you take the dual of this program, what you'll find is that the dual is asking you to construct a function p that has *zero* correlation with every degree d polynomial, but has large correlation with f. We call such a polynomial a "dual polynomial" for f, as one can think of p as a polynomial with no monomials of degree d or less. A dual polynomial for f is a 'witness' to the fact that f has large approximate degree. By strong LP duality, any approximate degree lower bound entails the existence of a good dual polynomial; it might just take a lot of work to find it! Both of our results -- our Omega(sqrt(n)) lower bound on the approximate degree of AND-OR, and our new proof of Paturi's result -- work by constructing explicit dual polynomials.

In the case of the AND-OR tree, we take a dual polynomial for AND, and a dual polynomial for OR, and we combine them in a certain way to get a dual polynomial for AND-OR. Our proof is a refinement of a technique of Sherstov -- Sherstov had already showed in a very general context the 'right' way to put together dual polynomials for simpler functions F, G to get a dual polynomial p for their 'block-composition' F(G, G, ..., G), but his earlier analysis could not handle the case that F=AND and G=OR. Prior work broke down because it was not known how to argue that p had high correlation with AND(OR, OR ..., OR). We manage to show this by exploiting some special properties of the AND function and the OR function (specifically, the key properties are that AND has low block sensitivity at all inputs except the All-True input, and dual polynomials for OR have 'one-sided error'. It's difficult to give much more detail than that in a blog post, so see the paper for the full discussion.)

We construct our dual polynomial for symmetric functions as follows. \v{S}palek , building on work of Szegedy, had already constructed a dual polynomial for the OR function i.e. a symmetric function with a 'jump' from +1 to -1 at the bottom Hamming layer of the hypercube. We give a relatively simple construction of a dual polynomial for the Majority function i.e. a symmetric function with a 'jump' from +1 to -1 near the middle Hamming layer of the hypercube. We then show how one can interpolate between these two 'extreme' constructions to give a dual polynomial for an arbitrary symmetric function f. Our final dual polynomial closely 'tracks' (in a sense that can be made precise by complementary slackness) an optimal approximating polynomial for symmetric f's, and we are planning on adding a section to the arxiv paper explaining this intuition.

Finally, we use LP duality to reprove some classical Markov-Bernstein type inequalities from approximation theory (not to be confused with Markov's Inequality or Bernstein's Inequality from elementary probability theory). Markov-Bernstein inequalities bound the derivative of a univariate polynomial q on real inputs in the interval [-1, 1] in terms of deg(q) and the maximum value q takes on this interval. Typically, these are the inequalities used to complete Step 2 in the outline of symmetrization arguments given above. We show how to formulate the problem of maximizing the derivative of a bounded polynomial as an LP, and give clean solutions to the dual LP that 'witness' various asymptotic versions of the Markov-Bernstein inequality. The connection between Markov-Bernstein inequalities and LPs has certainly been known before, but to our knowledge no one has proved these inequalities by analytically constructing dual solutions to these LPs, and we think our dual witnesses shed some new light on these well-known results.

Looking ahead, we're very optimistic than more open problems in the analysis of Boolean functions can be cracked using methods based on LP duality. Approximate degree is not the only measure of the complexity of a Boolean function f that can be characterized by a linear program: there's also polynomial threshold function (PTF) degree (i.e. the minimum degree of a polynomial that sign-represents f), PTF weight-degree tradeoffs (i.e. tradeoffs between the degree of a PTF for f vs. the size of its coefficients), and weight-degree tradeoffs for polynomials that approximate f pointwise to accuracy 1/3. All of these complexity measures have important applications in learning theory, communication complexity, and beyond, but remain poorly understood even in the context of simple function classes like AC^0.

This work was not done in a vacuum, and we are extremely grateful to Karthekeyan Chandrasekaran, Troy Lee, Sasha Sherstov, Robert \v{S}palek, Jon Ullman, Andrew Wan, and the anonymous ICALP reviewers for valuable comments and conversations! As well as to Ryan O'Donnell and Li-Yang Tan for suggesting the problem of constructing explicit dual witnesses for the LPs capturing Markov-type inequalities!

Congratulations to Mark Bun and Justin Thaler

Luca Aceto reported the paper awards for ICALP 2013.  But I find it necessary to give a special shout-out to Harvard Ph.D. students Mark Bun and Justin Thaler, who got the Track A Best Paper Award for their work:
Dual Lower Bounds for Approximate Degree and Markov-Bernstein Inequalities. 

I've asked Justin and he has agreed that he and Mark will write up a post about their paper that will appear in a few days.  But since it was announced I wanted to congratulate them now!  

This brings up an interesting question:  they won the Best Paper award, but not the Best Student Paper award.  Which makes some sense in one way -- if the "Best Paper award" dominates the "Best Student Paper award" (by definition), then there's not need to give the same paper two awards;  it has the nominally higher honor.  On the other hand, you might say that if they won the Best Paper award, by definition they should also have the Best Student Paper, so they should win both.  How do you think awards in that case should be distributed?

  

Guest Post by Mikkel Thorup : Results vs. Techniques

[Mikkel Thorup graciously has offered a guest blog post, discussing what he looks for in conference papers, and possible implications for refereeing.]


Results versus Techniques (Mikkel Thorup)
-------------------------

When I go to STOC/FOCS, I hope to see some great new results/theorems and some great new techniques/proofs. I want both, and in some wonderful cases, I even get both in the same paper. However, many of the greatest results may be in papers with less interesting techniques, or vice versa, and the worst thing is if they then get out-competed by papers achieving semi-strong results using semi-interesting techniques.

I am saying this because many refereeing forms have fields like "Submission's strengths" and "Submission's weakness" suggesting a balanced evaluation with pros and cons. A common case I see is that strong results get criticized for being obtained with too specialized techniques. However, if somebody needs to cite a theorem/result, then, typically, it doesn't matter if the underlying proof/technique is specialized.

I am arguing that if we want an interesting program, then we should not worry about weakness unless it renders a paper unacceptable (e.g., a bug). I am proposing a max-evaluation rather than a sum. What matters is if a paper has an exiting contribution.

I am myself mostly in the problem solving business where we believe that there are important computational problems. Sometimes general techniques work, but other cases require a deep understanding of the problem at hand leading to very problem specific techniques. Sometimes a tiny but subtle change in existing techniques can make a huge difference. Other times, as, e.g., for the 4-color theorem, it seems that what is needed is a huge mess. The point here is that if we want to succeed, then we have to be open to use whatever techniques it takes. Later there may be very interesting technical work simplifying and generalizing the techniques.

I am not saying that techniques are not important. I love papers introducing great new techniques, and I am more than happy to accept them on a technical basis. What I am saying is that if the main contribution of a paper is a result, then the techniques may play a more supporting role whose only obvious merit is that they prove something interesting.

It often turns out that the techniques developed to solve a central problem have other applications, but this may not be easy to guess up front based on a quick review. My experience is that referees are worse than random when it comers to guessing if a techniques will later prove useful in solving other problems.

My suggestion is positive refereeing where the overall score is based on the major contribution of a paper: the thing that would make people come to the talk, and use and cite the paper in the future. If a paper has something interesting to say, then it doesn't matter if other aspects of it are less interesting.

Upcoming Events

I've been asked to post for some upcoming events.

The Simons Institute will have a multi-day symposium on Visions of the Theory of Computing May 29-31, just before STOC.  Looks like a lot of big-name speaker.  More info here.

Ely Porat wants a lot of submission for SPIRE's 20th conference;  paper dealing is May 2.  Call for papers.  http://u.cs.biu.ac.il/~porately/spire2013/

Harvard E-Mail, Again

Yesterday was the long-awaited faculty meeting after the news that Harvard had examined the e-mail (metadata) of the Resident Deans looking for a leak of what was labelled confidential information. 

It was actually quite informative.  The most detailed account seems to appear (unsurprisingly) in Harvard Magazine.  To me, here are the highlights:

1)  Deans Smith and Hammonds gave unambiguous, complete apologies.
2)  There was a bit more e-mail searching than originally described;  in particular, there were some follow-on searches on the e-mail accounts of a Resident Dean that definitely did not go through the full process (specifically, Dean Smith doesn't appear to have been informed of them). 

I think the clear and unambiguous apologies were needed.  While I think I have been understanding throughout about why the administration felt the leak issue was important enough to merit e-mail searches, the fact was that Harvard's written policies were not followed.  That alone merits the apology.  Given that the previous apology offered by the administration was described by many as "half-hearted", it was important to have a full apology given at the meeting.  (Let's leave aside that it seemed a bit slow in coming.) 

It's a bit more disturbing that the institution doesn't seem to have a complete handle on process for searching electronic communications, but then again, it's a fairly new topic.  It's clear that this will be address on an ongoing basis, with an outside attorney investigating the past actions, and a Harvard committee being set up to examine the institutional policies regarding privacy of electronic communications.

Anyhow, seems like progress.

Harvard in the NCAA Tournament

Thank goodness it's Harvard's vacation week.
Otherwise, the rioting would have been unbearable. 

Cambridge Area Economics and Computation Day 2013

I was asked to advertise the upcoming Second Cambridge Area Economics and Computation Day (CAEC'13);
information is available at http://caec.seas.harvard.edu/

It's being co-chaired by Turing-award winner Silvio Micali and Harvard's own Yiling Chen.

The data/location is

Friday, April 26, 2013
Singleton Auditorium (46-3002)
Brain and Cognitive Sciences / McGovern Building, MIT
43 Vassar Street, Cambridge, MA 02139

Registration is free, but you have to register in advance.

Here's the call:

------------------------------
Call For Participation

A lot of research and business activity in the Cambridge/Boston area is engaged in economic and computational questions in regard to understanding and developing the economics of Internet activity. Examples of topics of interest include theoretical, modeling, algorithmic, and empirical work on electronic commerce, networked behavior, and social networks.
One of the main purposes of CAEC is to encourage collaboration between local researchers. Significant emphasis will be placed on a poster session and short talks. The overall structure of the day will involve four longer talks, by Daron Acemoglu, Yiling Chen, Matthew Jackson, and Silvio Micali, with a short talks session and a poster session over lunch, along with brief poster announcements.
Submissions

For short talks and posters, send an email to caec13@seas.harvard.edu by Thursday April 11, 2013, including a brief description of your work, along with an indication of a preference for the work to be presented as a short talk or a poster, or be considered for both. We will select a small number of short talks and put together a poster session.
Decisions about the program will be made by Monday April 15, 2013.
The suggested format for a short talk is (a) description of the problem, (b) statement of results, and (c) discussion of open research directions. There will be no time for setting up individual laptops for the short-talk session, instead we will have all presentations preloaded on a computer in the auditorium.
Registration

This event is free to participants, but PLEASE register by Monday April 22.
Breakfast and lunch will be provided for the event. There will be no cost to participants

Looking forward to seeing you there,
CAEC Steering Committee
Yiling Chen, Silvio Micali, Costis Daskalakis, Andrew Lo, and David Parkes 

Update: Andrew Auernheimer Watch

Well, this is going to give me nightmares for a while.

TechNewsWorld:

"Andrew Auernheimer, a hacker known as "Weev," was sentenced Monday to 41 months in prison for obtaining the personal data of more than 100,000 iPad owners from AT&T's publicly accessible website and sending the information to the media. The ruling immediately sparked an outcry from a digital rights group that claims the punishment does not fit the crime."

Wired:

Auernheimer and Spitler discovered that the site would leak e-mail addresses to anyone who provided it with a ICC-ID. So the two wrote a script – which they dubbed the “iPad 3G Account Slurper” — to mimic the behavior of numerous iPads contacting the web site in order to harvest the e-mail addresses of iPad users.
...
The two contacted the Gawker website to report the hole...
On Monday, following the announcement of his sentence, the Electronic Frontier Foundation announced that it had joined Auernheimer’s appellate team.

Slate:

"This is the third big CFAA-related case I’ve covered lately, the other two being those of Internet activist Aaron Swartz and Reuters deputy social media editor Matthew Keys. While the specifics of the charges in each case differ, all three illustrate the unfortunate plasticity of the CFAA, and how it can be shaped and contorted to cover almost any computing-related actions. (Did you fill out an NCAA bracket from your work computer today? Congratulations! Depending on your office’s computer use policies, you may have violated the CFAA!)"

Salon:

Heading into his sentencing hearing, “Weev” reportedly proclaimed, “I’m going to jail for doing arithmetic!” And in an earlier Twitter direct message conversation, the hacker told me what he believes to be the issues at stake in his case. He explained that the AT&T data he accessed was already “published” — the telecom company has admitted as much. “The government asserted that after the fact, they can declare a given access to data anyone makes public ‘unauthorized’ and have you thrown in prison,” he wrote.
...
As well as 41 months in jail, “Weev” has been sentenced to an added three years’ probation and must pay more than $73,000 in restitution to AT&T.
...
Auernheimer will appeal the federal court’s decision. On Monday, the Electronic Frontier Foundation announced that its attorneys would join his legal team. “Weev is facing more than three years in prison because he pointed out that a company failed to protect its users’ data, even though his actions didn’t harm anyone,” EFF senior staff attorney Marcia Hofmann said. “The punishments for computer crimes are seriously off-kilter, and Congress needs to fix them.”

But I Haven't Thought of You Lately At All....

I realize the times they-are-a-changing, and so maybe I shouldn't find this unusual, but...

Kristen Bell and Rob Thomas put up a Kickstarter campaign to raise 2 million to make a Veronica Mars movie.*  They gave the project 30 days to raise the money.

Done, in less than 24 hours. 

There have got to be huge lessons for e-commerce and marketing there.  I'm still trying to figure out what they are.  Maybe it's also just because I watched Amanda Palmer's TED talk recently, but the idea that you can now just go to the world and say, "Help support this project, that you will love" and it can happen, at a scale that I don't think was possible even ten years ago, feels very powerful.

Or perhaps I'm just happy that they're going to make a movie.  

* The first season of Veronica Mars remains one of the best things ever on television.  It's not just me saying that -- it was Joss Whedon and Kevin Smith.  Go buy the DVDs or something if you don't believe me.

 

Turing Award : Goldwasser and Micali

Got in the office this morning and heard about the latest Turing Award:  Congrats to Shafi Goldwasser and Silvio Micali!  (Press release.) 

A Second Life for our Groupon Paper

Groupon has been in the news lately -- something about their CEO leaving, I hear.  And Giorgos Zervas has been out giving talks.  (He's a great speaker.  You should invite him to your next related workshop or conference.)  The combination of these events seems to have given our Groupon work ([Byers Mitzenmacher Zervas] the newer arXiv paper, the older arXiv paper) a new life in the press.  I've seen it up on Forbes (by way of Rajiv Sethi's blog) and the Freakonomics blog, from which is seems to be being re-distributed through the Internet in the standard way.  Amusingly, the Freakonomics blog refers to it as a "new" paper (it's not), and nobody ever spells Giorgos's name right (this time, it's Servas instead of Zervas).

Since my name has been appearing in the press the past several days for less interesting reasons, it was nice to see it in the "news" for actual work. 

The Press

Harvard's name, obviously, still sells newspapers.  The "e-mailgate" scandal of the past few days has been appearing everywhere.  I was contacted multiple times from various news organizations over the weekend and yesterday -- the most contact from the press I've ever had.

I've chosen not to talk to them, not because I'm against the press, or because I'm naturally shy (I write a blog, after all), but because I don't feel I have anything much to say, other than what I'm saying here on the blog.  I don't have any additional facts other than what I'm getting from publicly available information.  I suppose the press is looking for my opinion, or more realistically some good quotes, but I'd rather not work through my opinions on this issue over the phone with a reporter.   

However, I guess I am "on the record":  reporters have quoted the blog.  While I'm not entirely comfortable with that, I suppose that's the way it is.  It's somewhat disappointing when you have to explain to your mother that yes, your name was in the New York Times, and no, you didn't do anything important.  

My first blog post on this was part of the narrative that the faculty was critical of the administration.  I noticed my second post has been interpreted that I've been "mollified" (a term used in at least one article) by the administration's explanation, which I do not think is accurate.  I'm glad the administration explained what happened from its point of view, and that it offered an apology (even if many interpret it as a half-hearted one);  I think the Harvard community was owed the explanation, and the Resident Deans were owed an apology. I acknowledge that the administration did a good thing to present that statement.  But I still say (quite clearly in the 2nd post) that I think the administration made a mistake, and there's clearly resulting damage they will have to deal with.  I can only imagine the Resident Deans are especially angry -- I don't believe I've seen a formal statement acknowledging that they are "faculty", but some of the initial speculation that the administration didn't feel they had to follow the "faculty policy" for them would, I surmise, be very insulting.  I don't believe the administration's statement addressed this point, but perhaps it has been addressed privately.  I also believe there's lasting damage to the trust between administration and faculty.  Harry Lewis has said he'll move much of his e-mail to a private gmail account;  I suspect many other faculty will do the same.  I also worry that this has damaged the nature of the relationships between Resident Deans and students, as Resident Deans are meant to be a "go-to" point for students experiencing a variety of problems.  Both sides, I suspect, will feel more wary of what they say to each other, if they are concerned the administration might be watching.  Those feelings may not be warranted, and the administration has tried to make that case in discussing the targeted methods they used when looking at the e-mail;  that doesn't mean they those feelings won't be there. 

I don't expect this issue to just go away without further discussion, at the upcoming Faculty Council and FAS Faculty meeting.  Some people still have questions as to how this came to pass.  Larger scale, this fits into a narrative I've heard multiple times the last few days, about the "corporatization" of universities in general and Harvard in particular.  It's an issue that many faculty naturally feel very strongly about.  This event may galvanize some reflection or future action on that front.

So again, I don't think this is over.  But I think the form it takes from this point may be much less interesting for the press.

There's also more positive news on "the press" front, but it's worthy of it's own, separate post, which will be right up. 

Mike Smith Explains, Apologizes

Given the hubbub that arose this weekend over the issue of Harvard examining the e-mail of some of the Resident Deans, it's gratifying (if unsurprising) that this morning Dean of the Faculty of Arts and Sciences Mike Smith has released a statement, which explains the incident from the administration's point of view.

My high-level take -- this has been blown out of proportion by the media, but it's certainly an issue the administration and faculty should discuss and work out together, so there's a common understanding.  

Here's my summary (with commentary), but of course you should go the source.
1)  It was apparent that an e-mail labeled confidential has made it's way out to the press.  While the e-mail itself might not have contained especially sensitive information, the administration was concerned.  Who knew what else might have or would be leaked?  There was also concern that other specific confidential information had leaked to the Crimson.

Commentary:  The Ad Board member, who apparently forwarded a note containing advice on how Resident Deans might advise their students with regard to the Gov 1030 case, made an understandable error in judgment.  It's best not to forward mail labeled confidential, even if it seems unimportant.  If they had just rephrased the advice in their own words, perhaps this would have been avoided.  The administration's concern is also understandable, given the student privacy issues around Ad Board proceedings.

2)  The administration says they asked the Ad Board members about the forwarded email, but "No one cam forward", and they were told this might cause an investigation.

Commentary:  Sounds good.  This was something that wasn't clear or known (at least by me) based on previous reports.

3)  The administration decided to do a targeted e-mail search, looking specifically at the "Resident Dean" accounts (and not the "individual Harvard email accounts");  they did not look at the email content, but just looked at the subject lines to determine if the email in question had been forwarded. 

Commentary:  While I pleased with the care with which they did the search so as to avoid violating the privacy of the Resident Deans, I don't think this care offers an excuse for not following the policy of informing the Resident Deans of the search.  I would still say a search on their email had been performed and, from my understanding of the policy, they should have been notified.  This is something the faculty and administration can and should discuss further.  (In particular, should searches be required in the future, I do think the administration should be encouraged to follow a similar process to specifically target the search.)

4)  After the forwarding had been found, the administration determined it was an inadvertent error, and no further action was taken. 

Commentary: Based on what I know, I agree, and think the administration took the right action.

5)  The administration made the decision not to inform the other Resident Deans, in order to protect the privacy of this Resident Dean and "[allow] the students cases being handled by this Resident Dean to move forward expeditiously."

And now here's the key statement to me in the statement:

"We understand that others may see the situation differently, and we apologize if any Resident Deans feel our communication of the investigation was insufficient."   

Commentary:  In the original story I saw at boston.com, Sharon Howell, senior Resident Dean, said the following:

“They don’t seem to think they’ve done anything wrong,” she said. “[I told them], if you want to repair this with the resident deans, it would make sense to talk about why you thought this was the right thing at the time, and apologize for not notifying us after the fact.”

I feel that the administration has done this.

I've seen postings in other places, where people are suggesting that this case demonstrates some sort of moral failing, and that someone somewhere should be dismissed.  I disagree.  In my opinion, a Resident Dean made an understandable (and I would argue small) error in judgment in forwarding an email marked confidential.  The administration was rightly concerned.  In my opinion, the administration made an error in judgment by not treating the Resident Deans as faculty and strictly following the Harvard policy by informing them that a search was being done as part of an investigation into the matter.  I'm not clear if they feel they made an error in judgment, but they have apologized.

The faculty now have a chance to discuss with the administration why they felt this was an error in judgment and how they'd like to see similar situations handled in the future.  I expect and hope these conversations will happen.  But I don't think that's the sort of exciting stuff that somehow gets into the New York Times. 

Harvard Spies on E-mails

A strange offshoot of the Harvard cheating scandal -- a report in the Boston Globe that Harvard administrators decided it was OK to search through Resident Dean e-mails when some information marked confidential had appeared to have been leaked.

Just some "vocabulary" for non-Harvard people.  Resident Deans are people who live in the Harvard dorms (what we call "houses") who help administrate the dorms and work with students.  For example, one of their jobs is to serve on the Administrative Board ("Ad Board"), which, as I've talked about, handles disciplinary cases, including the now infamous Gov 1030 cheating scandal.

Short version of the story:  it was noted that a confidential e-mail of the Ad Board had apparently leaked out to the Crimson.  The e-mail was pretty innocuous, but of course one doesn't want confidential e-mails leaking about.  So someone in the Harvard administration decided a bright idea was to check the e-mail accounts of the Resident Deans without telling them to see if they had leaked it.  (Apparently, one had.  Success!)  

Now, since Professors are the uber-people in the university*, we have rules to protect our e-mails from being searched except in extreme circumstances (and even then with notice), as the article points out.  I found it online after a bit of searching.  The first paragraph is:

"The Faculty of Arts and Sciences (FAS) provides the members of its faculty with computers, access to a computer network and computing services for business purposes, and it is expected that these resources will be used in an appropriate and professional manner. The FAS considers faculty email messages and other electronic documents stored on Harvard-owned computers to be confidential, and will not access them, except in the following circumstances."

(I was amused when the article noted "That FAS policy was drafted in part because some professors feared that previous president Lawrence Summers was monitoring their correspondence."  Summers really was quite unpopular with many faculty, and while I can't recall what specifically had some professors concerned about whether the administration could read or was reading our e-mail, I hope some of my colleagues can remind me in the comments -- I do recall that concern being around.)

So it would seem that the administration violated its own policy.  What ground do they have to stand on?  It seems like the only question is, are Resident Deans "faculty" under the policy?  They generally hold lecturer positions within the University, but they are also generally not tenure or tenure-track.  But as far as I can tell right now, they're faculty under any reasonable definition the university gives.

(The FAS Handbook described Lecturers as follows:
"A lectureship is a short-term, non-tenure-track position that is held by individuals who serve as course heads for courses that would otherwise be taught by tenure-track or tenured faculty. Lecturers must hold a Ph.D. All lecturer appointments must be based in a department or degree committee."
Sounds like faculty to me....) 

I was pleased to see CS faculty members Harry Lewis and Greg Morrisett quoted in the article, with both of them being pretty unhappy to hear about this.  Greg's comment that "a lot of faculty are going to be 'pissed off'" rings true to me.  (I assume I won't have to, but if nobody else makes a big deal about it -- hard to believe -- I would plan to bring it up at the next FAS faculty meeting.)

Also, it just seems silly.  They could have/should have first asked the Resident Deans if any of them had messed up, and let them know that Harvard considered the leak of Ad Board related confidential information a sufficiently big issue that they planned to search their e-mail if necessary.  My reading is the FAS policy would have allowed for that, according the following paragraph:

"Second, in extraordinary circumstances such as legal proceedings and internal Harvard investigations, faculty records may be accessed and copied by the administration. Such review requires the approval of the Dean of the FAS and the Office of the General Counsel. The faculty member is entitled to prior written notice that his or her records will be reviewed, unless circumstances make prior notification impossible, in which case the faculty member will be notified at the earliest possible opportunity."

One might quibble as to whether the e-mail leaked in this case corresponds to an "extraordinary circumstance", but if the administration had followed this policy, it would have been hard to argue with.  Why didn't they do this?  We (or at least I) don't know.  Heck, it could be that they forgot about this policy.  There are so many pages of policy around, it's hard to keep track of them all, as I've annoyingly found while serving as Area Dean.  Or they did remember the policy, but actively decided that Resident Deans weren't "faculty".  Or, they just really didn't care.  Lots of possibilities.  But the idea that nobody had the common sense to say, "Whoa!  Isn't it bad precedent to go looking through people's e-mail without going through other options or letting them know?" is the kind of thing that make faculty think they need to push back or even rein in the administrative side of the university.

So I expect to hear more on this issue.  

* Or at least we like to think so.  Though we can tell that the administrators increasingly think that they're the uber-people, and this can make us very unhappy in some circumstances.  Like this one.  

 UPDATE:  As is often the case these days, Harry Lewis and I are blogging in parallel on Harvard matters.  His insightful thoughts on this situation can be found here.  

Rankings Don't Matter (But)

Back when I started at Harvard, people would literally say things to me like, "I didn't know Harvard had computer science."

Now I could just point such people here.