Saturday, April 18, 2009

Putting Your Data on The Web



I just watched a talk by Tim Berners-Lee (who invented the web browser) that re-inspired me to put the data I collect on the web, so other people can have access to it.

I heard at one point that in psychology you were required to keep your data for five years after publication in case anybody else wanted to look at it. I think the expectation is that if someone asks you for your data, you are supposed to give it to them so they can duplicate your statistical analyses, or whatever.

However, now that data is, to a great degree, stored on and collected with computers, there's no reason to get rid of your data. And if you are going to give it away anyhow, why not just let anyone download it, rather than having to email you?

I run psychological experiments, and I want to start putting my data on the web. I keep a website (http://www.jimdavies.org/) and on it each paper I write gets its own web page (e.g., http://jimdavies.org/research/publications/ijcai/2001/davies2001.html). But now I'm thinking that I should have, linked off of these pages, the data collected. Not only that, everything I can put up there to help someone who wants to know more about what we did. For example, we collect data on computers, and we have to write programs to collect it. We use a piece of software called E-Prime to do it (other labs use "Superlab.") I also want to put the E-Prime source code up there, in case someone wants to replicate.

I also do AI work, and I think it's a good idea to put your AIs on the web too. I have not done this. I seem to rememeber being warned not to do this, when I was in graduate school, but right now I can't remember any reasons that outweighed the potential benefits that society might stand to gain from the access.

I plan to put my data on the web, and I hope all scientists reading this blog will do the same.

The other thing is that I think journals and conferences should maintain websites of the data they publish. This will force/encourage shared data, but also puts it safely in the hands of a big organization. When a scientist dies, who will maintain their website?

I might just write a letter to the cognitive science society suggesting this. 

Wednesday, April 15, 2009

Solving Rubik's Cube





I have had a Rubik's Cube since the 1980s and have never figured out how to solve it. At least now I can say I've solved it twice. I looked online and found this great video series that actually made it possible.

It taught me that it gets solved with repeating algorithms, which is something I didn't know. Now, with notes, I can solve it.

Next step is to memorize the algorithms, then figure out why they work. Sadly, I have no idea how I'm solving it.

Tuesday, April 14, 2009

My Students Are Adding To An Online Textbook


My teaching philosophy is not to waste student time. The last two times I taught my undergrad AI class I have instructed them to contribute to the online AI textbook at Wikibooks.org.

They had to describe a search method in AI. Now most of the main methods are covered. With hope, in ten years or so I won't even need them to buy a textbook.

Below are some of the wikibook chapters my students have made. Note that since anyone can edit the wikibook, they might have been changed since I posted this. 


A*

http://en.wikibooks.org/wiki/Artificial_Intelligence/Search/Heuristic_search/Astar_Search

Beam Search

http://en.wikibooks.org/wiki/Artificial_Intelligence/Search/Heuristic_search/Beam_search

Best-First Search

http://en.wikibooks.org/wiki/Artificial_Intelligence/Search/Heuristic_search/Best-first_search

 

Bidirectional Search

http://en.wikibooks.org/wiki/Artificial_Intelligence/Search/Heuristic_search/Bidirectional_Search

Breadth-First Search

http://en.wikibooks.org/wiki/Artificial_Intelligence/Search/Exhaustive_search/Breadth-first_search

 

Dijkstra's Algorithm

http://en.wikibooks.org/wiki/Artificial_Intelligence/Search/Dijkstra's_Algorithm

 

Finite State Automata

http://en.wikibooks.org/wiki/Artificial_Intelligence/Search/Exhaustive_search/Finite_state_automata

 

Hill Climbing

http://en.wikibooks.org/wiki/Artificial_Intelligence/Search/Iterative_Improvement/Hill_Climbing

 

Minimax Search

http://en.wikibooks.org/wiki/Artificial_Intelligence/Search/Heuristic_search/Minimax_Search

 

Recommender Systems

http://en.wikibooks.org/wiki/Artificial_Intelligence/Search/Recommender_Systems

Saturday, April 04, 2009

Who Gets Authorship?


I used to think it was really clear. I've been talking with my man Gabe Brostow, and it's not so much. Different fields conceive of it differently, and even within a field the borderline questions can get to you.

"Authorship" is who gets their name on a publication. "Authorship order" is who gets their name first, second, etc. 

First I'll talk about the differences between fields.

In most of the humanities, single-authored papers are the norm. Even if the primary author is a graduate student and has had extensive advice on the paper from the advisor, the student is the only author. This is the norm in philosophy, English, film studies, etc. In scientific field it's common to have the advisor's name on papers. The problem with the single-author culture is that it gives the advisors little external motivation to have and mentor graduate students. You can't put your grad students' papers on your CV. 

In computer science, the field I'm most familiar with, it's a matter of course that 1) your advisor mentors you a lot, and 2) that her* name goes on every paper. The principle investigator on a grant (the "PI") usually gets last authorship on every paper coming out of the lab. This means that although first author is the most prestigious, last author is the second most prestigious. She's usually the one who masterminds the whole lab's mission. 

Gabe tells me that in engineering, you don't actually talk about your research with your advisor. You talk about where to publish, who should be on your committee, things like that. He tells me that if an idea comes up while talking to your advisor, you can't use it in your dissertation. (!!) Nonetheless, the advisor gets her name on your publications. That seems very weird to me; since my working heuristic is that people get their names on the publications when they have made a significant intellectual contribution. 

I taught this heuristic to my undergrad cognitive science methodology class recently. One of the students ended up coming to my lab to work. We brainstormed the next step in the experiment, and she helped. She aske me later if she'd made an intellectual contribution and should she therefore get authorship. I liked her being up front about it. I told her that I was glad she told me that, because if she wants authorship we will make her work for it. She agreed to.

However, that project already has four authors on it. Is there any harm in adding another? Possibly. In some sense it dilutes the apparent contribution of everyone else. I also don't want resentment in the other authors who have done much more work. Finally, though it's not an issue with this many authors, if the next person is a third author then it changes the way it gets cited in many publications in the text. For example, if there's a single author it will look like (Davies, 2009). If there are two authors it will look like (Davies & Haigh, 2009.) If there are three it sometimes** gets shortened to (Davies, et al., 2009). So adding a third author reduces the name recognition being built for the second author. This is not a problem im many computer science publications, though, because citations occurr by number, as in [1]. You have to look at the references section to know who is being cited. In the humanities it's not a problem either, since the full citation is often in the footnotes.

I don't want everyone in my lab to have their name on every paper, especially since I have many students in my lab. I am thinking of implementing a general rule (which can have exceptions) that the lab meetings are for freely given advice. That is, if you are talking about your research in the lab meeting, and someone gives you advice, that does not necesarily mean that that person gets their name on a paper. You give them the same courtesy. If you want your name on the paper, you have to help write it or do something outside the general lab meeting to earn it-- and in that sense I'm pretty generous with authorship. I want to encourage a culture in my lab of collaboration and mutual help. I want people to feel free about coming to the lab meeting and asking questions and not worrying about having to add authors to their every time they do. Of course, there can be exceptions, such if someone in the lab meeting has a really, really great idea and should get credit for it. But even in this case I would hope the response would be to have the person join the project and help in other ways. 

I'm very interested in other thoughts on the topic. (If you leave a comment and don't sign your name it will appear as from "anonymous.")

Pictured: Somebody making a paper.

* The convention I use for the he/she thing is that if I'm referring to someone of higher status, like an advisor in an advisor/student relationship, it's a woman; else it's a man.

** In psychology (APA) style the first time you cite it all authors are listed. Subsequent citations get the et al. SIGGRAPH (The prestigious computer science conference of special interest group: graphics) does the et al. thing from start.

Friday, April 03, 2009

What Kind of Calligraphy Is That?



I went to the monthly meeting of the Ottawa Calligraphy Society tonight, as I try to do every month.  They encourage you to bring in some calligraphy to show, and provide a suggested word. This month it was "renewal." 

But as I sat waiting for the meeting to start, I did some more calligraphy with the pen in my pocket. I did the orange image at the top. See if you can figure it out what it is (I'll let you know at the end of the post.) 

Anyway, I put it on the gallery table with the "Renewal" piece above.  During the social time of the meeting, a woman approached me with the orange calligraphy and asked what kind it was. I told her that it was a kind I'd made up myself (which is true-- I adapted the techniques from a book on contemporary Chinese Calligraphy to English). She didn't like my answer. "But what kind is it?"

"It doesn't have a name. I made it up."

"But I mean, is it Italic, or Gothic...?"

I could not give her an answer that satisfied her, so she took my work to the president of the society. I saw them chat for about 10 seconds, and then she came back to me and handed me my calligraphy. She said, with confidence and satisfaction: "It's contemporary/cutting-edge."

The calligraphy is the crawl from Star Wars: A New Hope, which was printed on my folder.

Wednesday, April 01, 2009

The Disbelief in Evolution Should Not Be Protected As A Religious View


This is from the newspaper The Globe and Mail:
Canada's science minister, the man at the centre of the controversy over federal funding cuts to researchers, won't say if he believes in evolution.

"I'm not going to answer that question. I am a Christian, and I don't think anybody asking a question about my religion is appropriate," Gary Goodyear, the federal Minister of State for Science and Technology, said in an interview with The Globe and Mail.

A funding crunch, exacerbated by cuts in the January budget, has left many senior researchers across the county scrambling to find the money to continue their experiments.

Some have expressed concern that Mr. Goodyear, a chiropractor from Cambridge, Ont., is suspicious of science, perhaps because he is a creationist...

“It is the same as asking the gentleman, ‘Do you believe the world is flat?' and he doesn't answer on religious grounds,” said Dr. Alters. “Or gravity, or plate tectonics, or that the Earth goes around the sun.”

Minister won't confirm belief in evolution
is the link...

I've been reading this great book (Stanovich, 2004) with a great chapter on memes. A meme is an idea that gets spread through a population of people, somewhat like the way a gene gets spread through a population. He talks about how certain memes or sets of memes protect themselves from scrutiny. For example, by saying "everyone has a right to their opinion" is an implicit suggestion that the views being discussed should not be questioned, examined or justified. In the article above, Mr. Goodyear is asked about evolution. He probably doesn't believe in it, so what does he do? Rather than risk getting into a debate, he refers to his belief/non-belief in evolution as a religious issue.  And we all know that we can't question people's religious views? Right? So I guess we can't talk about evolution [insert sarcastic emoticon here].

With some of the absurd claims that religions have, it's no wonder they carry memes with them to protect them from rational thought. Faith is good, people have a right to believe whatever they want, etc. 

Pictured: A cartoon by Tom Schmal

References

Stanovich, K. (2004). The Robot’s Rebellion: Finding Meaning in the Age of Darwin. University of Chicago Press.