Joho the Blog » research

May 7, 2015

Facebook, filtering, polarization, and a flawed study?

Facebook researchers have published an article in Science, certainly one of the most prestigious peer-reviewed journals. It concludes (roughly) that Facebook’s filtering out of news from sources whose politics you disagree with does not cause as much polarization as some have thought.

Unfortunately, a set of researchers clustered around the Berkman Center think that the study’s methodology is deeply flawed, and that its conclusions badly misstate the actual findings. Here are three responses well worth reading:

Also see Eli Pariser‘s response.

1 Comment »

October 25, 2013

[dplafest] Advanced Research and the DPLA

I’m at a DPLAfest session. Jean Bauer (Digital Humanities Librarian, Brown U.), Jim Egan (English Prof, Brown), Kathryn Shaughnessy (Assoc. Prof, University Libraries, St. John’s U), and David Smth (Ass’t Prof CS, Northeastern).

Rather than liveblogging in this blog, I contributed to the collaboratively-written Google Doc designated for the session notes. It’s here.

Comments Off on [dplafest] Advanced Research and the DPLA

August 5, 2009

Media Cloud unclouds media

The NY Times has a terrific article about Media Cloud, a Berkman Center project (hats off to Ethan Zuckerman, Yochai Benkler, Hal Roberts, among others) that will let researchers track the actual movement of ideas through the mediasphere and blogosphere.

Data about concepts! What a concept!

[Tags: ]

Comments Off on Media Cloud unclouds media

February 21, 2008

[cyberinf] Cyber-enabled knowledge

Peter Freeman of the Washington Advisory Group introduces the first panel, on “Cyber-Enabled Knowledge”, by asking how the infrastructure can support the university’s essence as the creator, transmitter and preserver of knowledge. [As always, I’m paraphrasing, typing quickly, and undoubtedly getting things wrong.]

Guru Parulkar of Stanford says that we must build the cyberinf on the right foundation. That’s one that enables many layers. It requires supporting the end-to-end principle because that facilitates innovation. We should make the infrastructure programmable so that providers can give users empowering services. [Seems non end-to-end to me. But I think he’s talking about university infrastructure providers enabling experimental services, not having, say, Comcast build services.] It’s not enough to deploy vendors’ infrastructure on campuses. The CIO and researchers ought to get together on this.

Simon Porter, eScholarship Research Center, U of Melbourne, wonders what the world looks like when we can find about all the research going on in our university. We could manage portfolios of research under an overall university agenda. [Hmm. Possibly scary.] They could develop a data research plan. The university could plan its storage needs. The way the research is represented to the public will change: it won’t be left to the researchers to be the lead communicator about the project. There will be a single portal — like Amazon or eBay, perhaps — where you can find out about research. We will be able to evaluate research by the effect it has on other projects. Researchers will be able to cooperate more, especially if there are standards. Crystallographers have software that lets people annotate online models; this is promising.

Q: Simon and Guru both pointed to gaps between network engineering folks and the CIO. What’s blocking progress here?
Simon: It’s not a natural progression. You have to take a leap.
Guru: The infrastructure is so complex, there’s a reluctance to “muck it up.” But at Stanford there’s a lot of openness.
Peter: Market forces will bring about the healing of the gap.

Q: The infrastructure didn’t arrive on a gold cloud all at once. It’s built on standards. In a recent survey, only 30 universities (G7) had courses on standards. Standards aren’t taught or shared at universities.
Guru: I disagree with you completely. Universities should be doing research much before people think about standards.
A [we’ve been asked not to identify speakers without asking permission :( ]: The U’s are incredibly creative now. I believe the next thing will come primarily out of U’s. Things bubble up, and the standards follow after that.
Simon: Standards are fundamentally important for development of cyberinf.

Q: How do we change the research processes to take advantage of the new cyberinfrastructure. This is not a decision for the CIOs but for the college presidents, etc.
Peter: By acclamation, we agree.

Q: [me] Knowledge currently reflects the old infrastructure: You get published or not. Knowledge is binary, fenced in and managed. How will the new infrastructure change the nature of knowledge itself?
Simon: Especially with shared standards, research can be more open.
Peter: Simon has proposed a specific way to make available info about current reseach projects. That’s key to enabling cooperation and the development of standards.
Guru: The cyberinfs we deploy on our campuses should allow experimentation in networking, cooperation, etc. That type of infrastructure doesn’t exist because we haven’t been asking for that leel of programmability and flexibility.

Q (John Wilbanks): When we try to move from network standards to knowledge standards, we get into semantics. It’s hard to have enduring semantics because they change as research happens. We could have project-based standards and allow people to share what they mean about something, not just sharing the content. So we have to change the idea of standards. [Go John!

Q: Is it the U’s role to fund research into infrastructure? You can’t make a case to the provost unless you show some dollars coming from somewhere.
Guru: Yes, someone has to pay for it. Maybe vendor partnerships will help.
Simon: If it’s strategically important to the U, the U ought to do it.

A: I’m in bioinformatics. BTW, my U doesn’t teach any of the standards. Anyway, industry folks tell us we’re training students to be like you, not to be what we in industry need. E.g., not team players. How can we make more industry-academic partnerships?

A: There is something big going on that we don’t understand. We’re good at big networks, etc., but we don’t understand how to solve problems for small groups of collaborating domain scientists. Universities don’t just store, transfer and develop knowledge…

I direct one of the portals where project-based info can be shared. People keep asking what the incentive is for professors. Right now the reward structures are not geared towards publishing on the Internet. What can be done to fix the incentive system?
Simon: Making info available is always going to be a chore to researchers. But Facebook makes it possible for marketers to find info based on participation by users. We need something equivalent for researchers, surfacing info about projects without requiring additional work by the researchers.
Guru: If it’s a problem of aggregation? People are very eager to make their work public. Where is the disconnection?
Peter: It largely depends on the field.

A: I develop provenance metadata in my field. There are problems. Ontologies don’t exist yet. They require expertise in RDF as well as domain expertise, and that’s hard to find in the same person. The ontologies have to be developed internationally.

A: Maybe there are some Web 1.0 opportunities that haven’t been take advantage of yet. E.g., we could make available to any NSF researcher a Web page at the NSF site. That would also provide some authentication.
Simon: It’s not a web page. Every researcher needs a persistent identifier. [researcher or proejct??

A: Standards that have followed research experimentation and productization have been the most successful. E.g., Internet, LANs, the Web. The most spectacular standards failure was the OSI in the 1980s because they did it before they had the sw and the experiments.

A: At my [hardware infrastructure] company, we do a lot of rolling out of products internally that are not quite ready. We are probably more willing to risk failure than universities are. And we are seeing more demand for programmable infrastructure hardware.

I urge us to adopt a more expansive, active and empirically-grounded notion of infrastructure. We shouldn’t think of infrastructure as being primarily hardware. 1. The layer model encourages thinking of the hardware as the “real” stuff. 2. We need to be teaching our students the practices by which interoperability is made possible. The standards in ten years will be different, but the tensions and dynamics will stay roughly the same. 3. We should learn from previous attempts to build infrastructure.

A: Infrastructure is extremely important but that occurs in a multicultural environment that we should bear in mind. Second, it all comes down to open access. [Tags: ]


February 12, 2008

Harvard to vote on open access proposal

The NY Times reports that Harvard’s Faculty of Arts and Sciences will vote next week on a proposal that would require faculty to deposit a copy of their articles in an open access Harvard repository even as they submit those articles to academic journals.

I like this idea a lot. I only wish it went further. Faculty members will be allowed to opt-out of the requirement pretty much at will (as I understand it), which could vitiate it: If a prestigious journal accepts an article but only if it’s not been made openly available, faculty members may well decide it’s more important for their careers to be published in the journal. I would prefer to see the Harvard proposal paired with some form of official encouragement to tenure committees to look favorably upon faculty members who make their work widely and freely available.

Nothing is without drawbacks. A well-run, reliable, thorough peer-review system costs money. But there’s also an expense to funding peer review by limiting access to the work that makes it through the process. Likewise, while the current publication system directs our attention efficiently, but there’s a price to the very efficiency of such a system: innovation can arise from what looked liked inefficiencies. There’s value in the long tail of research.

If we were today building a system for evaluating scholarly research and for making it maximally available, we would not build anything like the current paper-based system. Well, we are building such a system. The Harvard proposal will, in my opinion, help.

Disclosure: I’m a fellow at the Berkman Center which is part of the Law School, not the Faculty of Arts and Sciences, and I’m not a faculty member in any case. Stuart Shieber, one of the sponsors of the proposal, is a director of the Center.) [Tags: ]