Tuesday, September 29, 2009

Blog Posts of the Day

A blog post worth reading is Mihai Patrascu's post on, essentially, coming in second, if only for the chance to play armchair psychologist and try to deconstruct Mihai based on his blog posts. Of particular interest to me was his reaction to being offered a job at UCSD as the second-choice candidate -- an offer which he turned down, and apparently would have taken if offered first.

This is interesting to me because this very issue came up in our last search (which I was leading), where we ended up making 6 offers (and got 3 acceptances). We (the hiring committee) recognized that we were making a rather significant request to have 6 simultaneous outstanding offers. We also recognized the dangers in trying to sequentialize these offers. First, there was the internal danger -- the complex discussions (we had such a great committee, we wouldn't have argued) we would have had to undertake to rank-order everyone we wanted to make an offer to. And second, there's the external danger that the candidate -- who will, of course, find out they were the "second choice" -- takes the ordering as a negative signal and becomes more inclined to take another offer. One can argue whether or not a candidate should take such an ordering as a signal, or whether such a reaction is a purely emotional response. (See Mihai's post, for example, and judge for yourself in that case.) But it was clear to us that, even if no such signal was intended, there was a high risk that would be the interpretation from the standpoint of the candidate.

Mihai's post provides a solid empirical data point that we were right to have this concern; it's something I will keep in mind (and if necessary point to) in future hiring discussions. I'm glad we were able to make 6 simultaneous offers, and give all of the candidates we made offers to the right signal.

Something not worth reading is Kevin Carey's article in the Chronicle of Higher Education, where he seems to be saying that Harvard should be doing more for undergraduates and in particular admitting more undergraduate students. It's so bad, I can't bring myself to link to it. Without judging the point of criticism, I've pointed out that his rant is pretty much devoid of an actual argument; if you care (and really, I'd suggest reading Mihai's stuff first!), you can see that I'm somehow now embroiled in an argument with him here and here.

45 comments:

Anonymous said...

Michael, I very much enjoyed your obliteration of Kevin Carey's argument. Thanks!

Anonymous said...

I guess it depends on how much you want arrogant colleagues.

D. Eppstein said...

To anonymous #2: I think some amount of arrogance is a good thing, as motivation to make the next result so over-the-top great that even the most obtuse of critics will recognize its greatness. I'd far rather have an arrogant colleague than an incompetent or mediocre one. But if one is going to throw a tantrum every time one comes in second-best, well, that's going to be a lot of time and energy throwing tantrums that could better have been spent doing great research. And as you hint, it is likely to grate on one's colleagues.

Anonymous said...

Well, I knew I was the second choice of the institution I ended up going to, and it never even occurred to me to take that as a negative. They wouldn't have made an offer just to insult me, after all---there's too much riding on it. And it's not like being second took away all my power. In fact, having another offer in hand and knowing they were waiting on someone else, I could apply some pressure to make them put a time limit on their first choice. And despite being second, I was still able to negotiate a better offer.

M said...

And as you hint, it is likely to grate on one's colleagues.

No kidding. People need to accept that, no matter how good they are or how hard they work, there's always someone who could be perceived as better at something important than them. Of the billions of humans on Earth, only one was Einstein, and even he won the Nobel Prize only once (primarily for the photoelectric effect, not for general or special relativity). Why should being second choice matter? Is anyone going to remember in a decade or two that you weren't the first choice? Heck, are most of your associates going to even know (barring your airing that information publicly)?

My first conference paper was rejected from a conference that had a rather low threshold in a field where journal papers were what mattered. When one of my teachers got wind of that, he looked and thought it should've gotten accepted, and eventually helped it to get accepted. I could have said, "Well, I'm not going; they didn't want me!" and that likely would have been the end for my conference-going career. But I didn't. And I've had many conference papers since. Likewise, I could have decided that if the topic scraped the bottom of the conference's barrel, clearly the corresponding journal didn't deserve my paper and surely would reject it out of hand. But I didn't conclude that, and the resulting paper had fairly smooth sailing to publication.

In other words, an initial rejection by the powers that be doesn't mean that something would be a bad match. This isn't like being a jilted lover. If it's the best job available for you, there's no reason why it should matter that you weren't initially perceived as the best for them.

I also can't help but wonder: Was UCSD his first choice prior to knowing who would extend offers? If not, there's a bit of hypocrisy to all this.

Anonymous said...

@M: UCSD was not his first choice: he was disappointed that none of the "top schools" interviewed him, and clearly he does not rank UCSD up there.

Top schools did interview Costis, and hired him, so as clearly, there is a wide agreement that Costis is more deserving of an offer (or a wide delusion that he is not, depending on the point of view).

If you read the post below his sour grape post, he talks about "labs vs academia" and rules out California on the grounds of the state "keeping medical doctors away." So again, very clearly, he had no intent to join UCSD, and uses the "2nd best" excuse to broadcast that he DID get the offer but was too cool to accept it. Or something.

Now as to what this should mean with respect to Harvard's admission process: nothing. Unless they want to pamper whiny crybabies.

Anonymous said...

I'd far rather have an arrogant colleague than an incompetent or mediocre one.

I don't see your point. What about incompetent, mediocre and arrogant?

You seem to think of arrogance as an expression of some objective greatness. I don't see an evidence for that.

There are great scientist who are humble and many more not-so-great scientist that are arrogant.
Arrogance is certainly a weakness. Not a strength.

On the other hand, weak people might be more motivated than others, in order to compensate for their weaknesses. Their arrogance is then a symptom of low self-esteem.

Andrew said...

As a guy who struggled through years of rejections to finally get one offer (I blame the tough mid-2000s market ... yeah ... that's it), getting an offer but being second choice strikes me as an awfully luxurious thing to complain about.

Anonymous said...

If Terry Tao becomes very jumpy, that is acceptable. If a mathematician has a few of Annals paper and becomes jumpy, it is tolerable. But Mihai, for what? A few S/F papers that only a few (< 5 I guess) people in TCS will read ? This is a primary example why the conference system in TCS is bad, because it can boost a nice kid's ego to a ridiculous level.

Anonymous said...

There is no credibility to the idea that a coming second reflects badly on the institution making the offer. There is simply no rational behind declining the UCSD offer. A position is not an award but an opportunity. Such a reaction is completely emotional even if the candidate himself claims otherwise.

Anonymous said...

With your secret methods revealed, unless I am the only offer Harvard makes for my field, I will know that I may, despite the offer, have been the distant second choice and will have to refuse as a matter of principle!

Anonymous said...

Didn't these papers "that only a few (< 5 I guess) people in TCS will read" solve very difficult open problems that were open for years? Doesn't that count for something? Is this a popularity contest or science?

Anonymous said...

Some problems are difficult,
because they are really difficult;
Some problems are difficult,
because < 5 people knew about
it and thought about it.

Anonymous said...

Mihai's post provides a solid empirical data point that we were right to have this concern

Michael -- IMHO Mihai's post does not provide empirical evidence of anything. As his post clearly shows, the guy is a crank. And that is all I think you should read from his post :-) Any reasonable candidate would understand the realities of the job market and behave in a more rational way.

Michael Mitzenmacher said...

Anon 11: You made me laugh. (Your advice shall inspire graduate students everywhere, I'm sure.)

Anon 14 (and others) : Given the comments here, I feel that first I should say I personally offer no judgment regarding Mihai's decision, and did not intend to denigrate Mihai in any way. I brought this post up for the reason I described -- the interesting issue of the potential effect of being the "second choice" can have on a candidate when being given an offer. The consensus I take from the comments (including Anon 14) is that that signal is meaningless and should be ignored, which is interesting and useful advice for future job-seekers. I have to admit, as a fresh PhD student, I would have had trouble looking at it that way, but perhaps I was simply less rational way back then.

Anonymous said...

a job at UCSD as the second-choice candidate -- an offer which he turned down, and apparently would have taken if offered first.

I'm not sure Mihai actually says that. In fact, given his post and responses to comments, I seriously doubt if he would have taken the job either way.

The consensus I take from the comments (including Anon 14) is that that signal is meaningless and should be ignored, which is interesting and useful advice for future job-seekers. I have to admit, as a fresh PhD student, I would have had trouble looking at it that way, but perhaps I was simply less rational way back then.
While on aggregate the signal may mean something, I am not sure there is anything new to be learnt from the fact that univ X rates A above B, once you have learnt that tens of other universities (conclusively) have the same relative ranking. In fact if univ X were to give an offer to B over A, a more likely reason would be that they realize their chances of getting A are low. Am I missing something?

Anonymous said...

Some problems are difficult,
because < 5 people knew about
it and thought about it.


You're being grossly dishonest here. Range counting, predecessor, voronoi diagrams, dynamic optimality, planar point location, nearest neighbor, partial sums, connectivity. Basically every computer science undergraduate who takes an algorithms course is exposed to most if not all of these problems. Don't let your personal feelings for the man cause you to misrepresent facts.

Anonymous said...

To Anon 17:

"Range counting, predecessor, voronoi diagrams, dynamic optimality, planar point location, nearest neighbor, partial sums, connectivity."

I got my phd in computer science theory, but I have never heard about the problems you listed. ( I am serious here.) I guess that this makes me a good example of having top conference papers but apparently not as good as an undergraduate. So having papers in top conferences does not mean too much about the author's qualification, right? Then why does Mihai make a big deal about his S/F papers?

Stefan Savage said...

Without touching on the particular case (the penalty of knowledge is that you can't speculate anymore :-) I think its important for everyone to realize that getting a job -- just like getting a paper published -- hinges on a great many factors only a subset of which are under your own control. For each candidate, understand that not only are there a number of individual factors (technical, social, fit, etc) but each of these is perceived and weighted differently by all of those involved in the decision making (its important to remember that in academia the majority of those making the decision are frequently NOT in your area). As well, there are almost always apples-to-oranges comparisons that play critical roles (how does one compare a candidate in TCS to one in graphics or one in HCI or one in systems?) and the politics associated with those comparisons do as well (how does one schedule scarce resources?); not to mention both the persistent and transient external political factors imposed by higher level decision makers (this wouldn't be a good year to try to get a job anywhere in UC for example). I've long felt that its a fallacy that there exists a fine-grained Platonic ideal of "goodness" for researchers (so too for papers), but its an even bigger fallacy is to expect that decision makers would abide by such a scale even if it existed. In my experience, job offers are job offers, just as paper acceptances are paper acceptances. Trying to analyze such results at a finer or deeper scale is unlikely to reveal many useful truths.

D. Eppstein said...

I got my phd in computer science theory, but I have never heard about the problems you listed. ( I am serious here.)

Ok, you and the faculty responsible for the program you got your Ph.D. in should be seriously embarrassed. Especially re the predecessor problem: it's the one binary search trees solve. But the others are also very important data structure ideas.

Anonymous said...

I try everything I can to take as many math courses as possible and to avoid computer science courses. Should I feel embarrassed for being called a computer science phd? I guess so.

D. Eppstein said...

Taking math courses (or otherwise learning the material): good. But avoiding learning basic computer science (and getting a computer science Ph.D. anyway): not good.

Fortunately you can still learn. Go read Demaine's lecture notes on data structures. Or if that's too much jumping into the deep end, start with the data structure sections of CLRS, but don't stop with them: data structures are CLRS' weakest point.

Anonymous said...

continue from the 21

The funny thing is, I always feel that I have enough computer science and do not have enough math when doing TCS.

D. Eppstein said...

The funny thing is, I always feel that I have enough computer science and do not have enough math when doing TCS.

To play Mihai for a moment, doesn't this give you a nagging feeling that you (along with the rest of the STOC/FOCS community) are letting yourself become a second-rate mathematician instead of a first-rate computer scientist?

I'm not trying to put down math. Math is good. I use lots of math all the time. I love math, for its own sake and for its applications to CS. Many of my papers are really math with little or no computer science content. But we should be developing our own math, not just being an import/export industry from the real mathematicians, and to do that requires not just mathematical sophistication but also a deep understanding of the computational structures we're trying to study.

Anonymous said...

""Range counting, predecessor, voronoi diagrams, dynamic optimality, planar point location, nearest neighbor, partial sums, connectivity."

I got my phd in computer science theory, but I have never heard about the problems you listed. "

The only thing more amazing than your ignorance, immaturity, and stupidity is how polite and composed D. Eppstein was in his response to you.

What kind of CS PhD brags about avoiding fundamental CS education?

I mean it might be acceptable at some level to not be familiar with certain topics however fundamental they may be, especially if you are a specialist; but really.. you're proud of this?


I am not really offended by anything you said. I just feel disgusted that mental midgets like you exist within the community. But then again, you're probably a nobody...

Anonymous said...

Yes I am a nobody, but I am a nobody but not cocky and jumpy in public, is it a little better than Mihai, which is nobody in my view, but acts very jumpy and cocky?

Anyway, I think to be jumpy and cocky, you need results like "prime is in P", which gain respect from general math community, or at least you need sth like recent result of Braverman,
that every theory blogs mention. If you do not have things like these, you are a nobody, just try to calm down and work. If the threshold is too low, the TCS is going to look like a kinder garden.

Sorry that my post may offer you in some way. My deepest apologies.

Anonymous said...

Anyway, I think to be jumpy and cocky, you need results like "prime is in P", which gain respect from general math community, or at least you need sth like recent result of Braverman,
that every theory blogs mention.


I have my PhD in theoretical CS, but have never heard of Braverman, and don't understand what "prime is in P" means -- never heard of that problem. But at least I know what connectivity is.

Anonymous said...

"ignorance"? "immaturity"?? "stupidity" ???
"disgusted"?! "mental midgets"!!! "nobody" (?)



Are you sure you feel OK?

Anonymous said...

"Anyway, I think to be jumpy and cocky, you need results like "prime is in P", which gain respect from general math community, "

Why would you assume gaining the respect of the general math community is a priority or even a goal?

Anonymous said...

By the way, I'm not sure "prime is in P" is the best example of a result with respect from the general math community.

M said...

When I Google "predecessor problem," click on the first result, and scroll down, the name "David Eppstein" immediately jumps out. So, with all due respect, I think the professor's perception of the importance of knowing the problem might be somewhat biased. My undergrad was in CS and I took four algorithms courses (2 undergrad, 2 grad, all at two of the top five rated CS schools in the country). I read the algorithms book for my first course from front to back. Most of my papers, though not in CS, have an algorithmic component. Still, I seriously doubt that "every computer science undergraduate who takes an algorithms course is exposed to most if not all of these problems." Maybe it's the case at UCI, which I've visited and have nothing but respect for, but not at other schools which are such as good. In fact, most of the topics listed I learned in other classes, not TCS classes.

Anonymous said...

I want to apologize again for my post. To make anybody other than Mihai mad is not my intention (actually I do not even want to make Mihai mad). I think that there is a misunderstanding when I said that a problem <5 people care. There is nothing negative about it. Most of academic problems are like that, including the problem I am working on now. That is why we are called ivory tower, right? And you donot need to solve pnp problem to get your papers into S/F, trust me.

--Nobody

Anonymous said...

"I have my PhD in theoretical CS, but have never heard of Braverman, and don't understand what "prime is in P" means -- never heard of that problem. But at least I know what connectivity is."

Vow, I envy you because you can study Mihai's great papers now. Keep on the good work.

Anonymous said...

doesn't this give you a nagging feeling that you (along with the rest of the STOC/FOCS community) are letting yourself become a second-rate mathematician instead of a first-rate computer scientist?


A good case can be made that most concepts
(including data-structures in TCS) are trivial once one has sufficient mathematical maturity and experience (say dealing with abstract structures).
Thus, its good just to take math graduate courses. The TCS concepts can be picked up later without too much effort. In this way one can be first rate mathematician working in TCS which really is a branch of pure math.

Anonymous said...

Hahahaha! So because you don't know the listed problems, and consequently the list of people who has worked on them.. this implies the problems are unheard of and less than 5 people work on them.. I would really suggest getting mental help at this point because your self-centeredness is off the charts.

Most of the problems listed (Range counting, predecessor, voronoi diagrams, dynamic optimality, planar point location, nearest neighbor, partial sums, connectivity) are fundamental, important, and/or have a lot of people working on them.

I don't like Mihai's antics for many reasons but you attacking his body of work is laughable at best.

What's even dumber is you downplaying the importance of problems he works on (like putting down predecessor) just to attack Mihai. Mihai didn't come up with the problem. So if you hate Alex Andoni, you're gonna say the nearest neighbors problem is some obscure problem nobody works on?

Seriously.. take it out on a bottle of anti-depressants, not theory blogs.

Anonymous said...

"A good case can be made that most concepts
(including data-structures in TCS) are trivial once one has sufficient mathematical maturity and experience (say dealing with abstract structures).
Thus, its good just to take math graduate courses. The TCS concepts can be picked up later without too much effort. In this way one can be first rate mathematician working in TCS which really is a branch of pure math."

yea.. sure.. whatever helps you sleep at night...

David Andersen said...

M, and earlier anon: I'll weigh in briefly on this, as a non-theory person who likes theory. I think that the claim that "most undergraduates" will have encountered these concepts is probably a reach -- as someone who took the graphics track (hey, I was at Utah - it was hard to resist) as an option instead of the theory track, I wasn't exposed to too many of them. But the introductory graduate algorithms course at MIT hit more than half of them.

It's clear that Mihai's post violated some widely-assumed social contract (I'll confess that I found it offensive), but in Mihai's defense, he's done an impressive body of technical work. There's something to be said on both sides of this debate for representing (seeing?) the world the way it is. There's nothing particularly contradictory or novel in being technically talented while having a talent for, ahh, irking others. :)

Anonymous said...

I wonder how much jealousy has to do with all of this. Mihai's work is very impressive, and I think way beyond the ability of most people. So it's quite natural to be jealous of him (I know I am), and therefore to pounce on him given the chance.

If you think about it objectively, there is something disturbing about the fact that one of the strongest graduates was not even interviewed at top schools.

Is it really a popularity contest, and not about science?

Anonymous said...

There's something to be said on both sides of this debate for representing (seeing?) the world the way it is.

Mihai's work is very impressive, and I think way beyond the ability of most people.[...]If you think about it objectively, there is something disturbing about the fact that one of the strongest graduates was not even interviewed at top schools.

Is it really a popularity contest, and not about science?



I am afraid I disagree with some of the implications here.

First of all, I agree that Mihai's work is technically solid, and on important problems.

But Mihai's post basically claims that his work is superior to that of his competitors. He suggests that the fact that a school that rates Costis above him in their ranking "cast(s) some doubt in [his] mind regarding [their] long-term strategy for becoming one of the top research departments in the theory of computation."

Putting aside the issue of what other factors ucsd may be taking into account other than raw research power, I strongly disagree with the suggestion that Costis' work, or research ability is weaker than Mihai's. In fact, I think anyone familiar with the work of both of them (excluding Mihai, but including Mihai's letter writers) would consider it absurd to make such a claim.

I don't think doing research in a popular area automatically makes it weaker. Having an impact on areas outside of CS does not automatically make the CS contribution weaker. Both Costis and Mihai solved long open problems. Both were technically challenging. Just because one of them was on questions few people have thought about in a decade or two does not make it automatically stronger.

So No. Thinking that you are better than others when the rest of the world seems to disagree is not "representing (seeing?) the world the way it is."

And No. It is not a popularity contest, but something popular can also be good science.

As a side note, given two candidates with comparable scientific contribution, I think anyone would prefer someone who is pleasant, someone who can understand and appreciate what his/her colleagues do, someone who is capable of working with other people, someone who will not have be handled with care lest he get upset.

David Andersen said...

Prev anon:

Please don't conflate my comments with the anonymous poster below mine. :) I wasn't trying to argue the relative merits of Mihai's work vs. that of others; I'm certainly not qualified to judge in that area. I was simply trying to say that someone's personality characteristics and the merit of their technical work are independent variables. I don't think it's intellectually honest to downplay someone's technical contribution because of the personality aspects.

But that doesn't have anything to do with the decision about who you want as a colleague. I agree completely with you on that note, and I believe it's both fair and rational to consider a job candidate in toto -- after all, you're going to have to work with them for potentially many years.

Anonymous said...

"Seriously.. take it out on a bottle of anti-depressants, not theory blogs."

Hi Buddy, not everybody who disagree with Mihai's greatness needs anti-depressants.

Suppose that you work in cryptography and know very little about data structures, then suddenly there is a guy named Mihai who you never hear before jumping out claiming he is the biggest star ever produced in TCS, what is your natural reaction?

I believe that most problems in academic worlds, especially in math, are studied by < 5 people. That is why I find academy so attractive. You feel that you are one of the few people in the world who really care about the problem. The problem somehow becomes your kid. I may be wrong in Mihai's case, but please allow people to have a different opinion.

My last suggestion to you:
Do not get too excited about some anonymous comments from "nobody". It is not worthy it, OK?

Anonymous said...

"Range counting, predecessor, voronoi diagrams, dynamic optimality, planar point location, nearest neighbor, partial sums, connectivity."...

Just to follow up on this thread: I am a theory faculty member at a top-20 school, and went to grad school at a (different) top-20 school. The Intro Grad Algorithms class at both does not cover most of these topics, and I can see how a theory person doing something other than algorithms might not be aware of these topics.

Anonymous said...


Just to follow up on this thread: I am a theory faculty member at a top-20 school, and went to grad school at a (different) top-20 school. The Intro Grad Algorithms class at both does not cover most of these topics,


Likely they were covered but under different names. The predecessor problem is usually called "binary search trees", dynamic optimality is called splay trees, connectivity is called BFS/DFS.

Anonymous said...

A lot of the anonymous comments defending Mihai sound like Mihai himself. Maybe it's time people stop responding to him - it's a big timesink. It's okay for a field to have one crazy person, but it's a problem if the whole field gets involved in his craziness on a regular basis.

Anonymous said...

To the anonymous before this: do you really believe that it's Mihai who is posting all these? I don't and I am not Mihai.. I am just a grad student at some "not top 5" TCS schools, and I find it completely dumb to think that Mihai is posting all these comments. What I conclude is that you're a Mihai-hater and need help.

Mihai is just different. It is not surprising he's a bit controversial and he's certainly not the first person who is a bit different and is attracting some controversy.