School of
Information
Previously School of Library & Information Studies
Friday Afternoon Seminar: Summaries.
296a-1 Seminar: Information Access, Spring 2024.
Fridays 3-5. Everyone interested is welcome.
Details are added as they become available.
In person, with also Zoom -- unless indicated otherwise. Campus policy requires
all Zoom participants to sign into a Zoom account prior to joining
meetings hosted by UC Berkeley. Face mask recommended but not required
for in-person attendance. Zoom sessions are not recorded.
Zoom link available only at
the School's Seminar event listing: www.ischool.berkeley.edu/events/ias.
Schedule. Weekly
mailing list.
Aug 30: *No Seminar meeting.*
Sep 6: Michael BUCKLAND: Relevance, Expressions and Context.
Last semester I presented a narrow definition of
the problematic notion of "relevance" and a descriptive "linear" model.
Extending that work leads to a comparable model of expressions
(signs, documents). Relating the two models yields descriptive
models of scholarship, persuasion, and both personal and
communal knowledge. It also invites attention to the concept of
"context". This is joint work with Wayne de Fremery. Our
published definition
and linear model of relevance has been published as: Relevance and creativity —
A linear model.
Sep 13: Eric MEYER, Dean: Knowledge Machines and Beyond: Digital
Transformations of Scholarship.
In 2015, Eric Meyer and Ralph Schroeder published Knowledge
Machines: Digital Transformations of the Sciences and Humanities. The
book was “an examination of the ways that digital and networked technologies
have fundamentally changed research practices in disciplines from astronomy
to literary analysis.” The book summarizes the e-Research and e-Science era
in the U.K., U.S., and elsewhere which dominated from roughly 2000-2010,
and the shift toward ‘big data’ in the first half of the 2010s. This talk
will present a summary of that work, but also reflect on more recent changes
wrought by machine learning, LLMs, and other advanced computational techniques
that continue to shape scholarship. Participants wishing to read the book can access the e-book through the MIT Press Backfile via the UC Berkeley libraries.
Eric Meyer joined the School as Dean and professor
on August 1. More at New Dean.
Sep 20: Jeffery HART: Hollywood Goes Digital: How Technology is
Changing the Movie Industry.
A number of digital technologies have reshaped the
movie industry: e.g., DVDs, digital cameras, digital post-production,
and online streaming services via the Internet. The COVID pandemic
resulted in a major shift away from going to theaters toward viewing
content at home. Digital libraries of films have become a major
source of revenues for what used to be called movie studios. What
are the implications of these changes for the future of the industry?
Jeffrey A. Hart is an emeritus professor of
political science at Indiana University in Bloomington, Indiana. His
main area of specialization is international political economy. Most
of his research has been on the politics of international economic
competitiveness in the advanced industrial nations. Between 1996 and
2001, he collaborated with Stefanie Lenway and Tom Murtha on the global
flat panel display industry, supported by a grant from the Alfred P.
Sloan Foundation. In 2001, he completed a project on globalization in
collaboration with Aseem Prakash that resulted in the publication of
three edited volumes. In 2004, he published a book on the politics of
high definition television (HDTV). He co-authored three editions of a
textbook with Joan Edelman Spero (The Politics of International
Economic Relations). His most recent book was Essays
on the History and Politics of the Internet (2023).
Sep 27: Combined with the School's 106th Birthday Celebration in South Hall
202.
In 1918, UC Berkeley began a full-time program in
library science, the direct ancestor of the present school.
See 106th Birthday Celebration.
Oct 4: Letters from Tokyo, 1950-1956 and Students' Reports.
Michael BUCKLAND: Robert Gitler's Letters from Tokyo,
1950-1956.
In 1950 the American Library Association sent
our alumnus Robert Gitler '31 to
Japan on a vague mission to establish a library school. With a modest
grant and great dedication he rather improbably succeeded. Today his
Japan Library School flourishes as an iSchool at Keio University, Tokyo.
Last year two trunks bought at a yard sale were found to contain over
900 letters from Tokyo to his mother which candidly described his
daily advantures.
At the end of the Seminar, if time permits, I will also
report on problems and progress replacing
the former ASIST "Pioneers" webpages with a wikipedia-format
historical biographical directory with international coverage.
4:15 - 4:45 pm: Students' Progress Reports:
Kui CAI: Impact of Climate Change on Food Insecurity and Micronutrient Deficiencies in the Democratic Republic of Congo.
This report examines how climate change intensifies food insecurity and micronutrient deficiencies in the Democratic Republic of Congo. It discusses soil degradation, disrupted agricultural systems, and limited healthcare infrastructure, highlighting the urgent need for sustainable interventions to combat malnutrition and health challenges.
Michael WANG: Error and Robustness of AI Text Detection
Methods.
As the capabilities of state-of-the-art LLMs become more and more powerful, the challenge of detecting AI generated text also becomes increasingly difficult. Research in this field has yielded several promising methods, while also raising concerns about error and robustness. These concerns must be addressed and resolved to enable practical and reliable applications of AI text detection.
Oct 11: Paul DUGUID: Branding Information.
Economics has contributed significantly to the spread of
modern notions of "information". To understand both how the field
deployed ideas of information and how this deployment may have
influenced the use of the concept more generally, this talk focuses on
one topic, intellectual property, looking in particular at economists'
accounts of trademarks. The talk investigates assumptions implicit
in celebrated economic accounts of brands, and concludes that the
fields's information-based assumptions significantly underestimate
the complexity of both brands and information. Such an exploration
is intended to help us understand more generally how we might address
important questions around the authenticity and reliability of information.
Oct 18: Deirdre MULLIGAN: New Directions in Tech Governance.
While congress continued to study the risks and
possibilities of AI, the Biden-Harris Administration took bold action:
driving agency actions to address risks to civil rights, equity,
competition, economic opportunity, and national security; establishing
new federal guidance to guide agency development, use, and procurement
of AI, and a new AI Safety Institute; and boosting the capacity of
government to use and regulate AI by bringing in new tech and tech-related
talent and driving public and private investments in the growing public
interest tech ecosystem. Drawing on my eighteen months of service in
the White House Office of Science and Technology Policy during the Biden-Harris Administration, I will describe key AI initiatives and, drawing on my
prior research, describe how these initiatives pave the way for the
government to purposefully use technology to embed values or set
policy—-what I call "governance-by-design"—-in a manner that supports
fundamental democratic governance norms of intentional, deliberative,
participatory, and expert public decision making, free from capture or
caprice, and centers the public's rights and safety over private
interests. Lastly, Mulligan will explain why these new directions in
tech governance make growing the cultural and institutional supports
for public service across the computing field should be an important
shared national priority.
Deidre Mulligan recently spent eighteen months as
Principal Deputy U.S. Chief Technology Officer in the White House Office of Science and Technology Policy.
Oct 25: Clifford LYNCH: Stewardship and Developments in AI.
This discussion will explore three separate but
inter-related areas where developments in AI technologies, particularly
generative AI systems, interact with digital preservation and
stewardship in complex ways. This is very early and speculative work.
1. Definitions of "open" AI components and systems. While
there is a reasonable consensus on the definitions and practices
associated with open source for traditional computer programs (and
an understanding of how open source interacts with preservation and
stewardship) a similar definition for AI has proven difficult and
controversial, and there's been limited examination of preservation
interactions.
2. Goals of stewardship programs for generative AI. This
follows on a discussion from last semester's seminar. It is not clear
what aspects and impacts of the behavior of generative AI systems a
stewardship program is trying to capture. I'll summarize last semester's
discussion and conclusions and seek to test and perhaps re-validate them.
3. Implications of chatbots designed to mimic individuals.
There have been a series of (perhaps over-hyped) interesting developments
here; I will summarize a few of these, including so-called "griefbots",
and open a discussion of the potential stewardship implications of these
computational objects or services.
Nov 1: Michael WANG: Benchmarking Error and Robustness of AI Text
Detection Methods. Brief progress report.
Nick MERRILL, Center for Long-Term Cybersecurity:
Competition "Inside" the Cloud?
A seldom-articulated (but commonly held) hypothesis is that
cloud services are inherently monopolistic: they enjoy tremendous economies
of scale, the benefits of which they pass on to consumers. If true, this
hypothesis would predict competition (i.e., breaking up big cloud providers)
to be an inappropriate remedy: since competition necessarily decreases the
scale of the largest competitors, it would also decrease their economies
of scale, and consumers would suffer in the form of higher prices, lower
quality, or both. This talk will present evidence to the contrary. We find
evidence that smaller providers can produce services competitive with
larger ones, at least in certain markets. Our results imply that the
market for cloud services is likely many markets and that
competition may indeed prove to be an effective remedy to concentration
for some of these markets. We seek to recover empirical means for
discovering these market segments.
Nick Merrill directs the Daylight
Lab at the UC Berkeley Center for Long-Term Cybersecurity. For Nick see Nick Merrill.
Nov 8: Kora CAI: Maternal Nutrition and Child Health in the Democratic
Republic of Congo: Trends and Key Indicators (2007, 2013 and 2014).
Brief progress report.
Guenther WAIBEL, California Digital Library: AI and Open Access.
The university generates (scholars) and stewards (libraries)
vast amounts of peer-reviewed scholarly content, and AI companies are
increasingly motivated to incorporate this content into their products,
including through agreements with academic publishers. Taking author
consternation about these agreements as its point of departure, this
talk will examine the intersection of AI, scholarly publishing and open
access, and explore the implications of different publishing models on
the university’s mission, self-interest and the public good.
Günter Waibel is Associate Vice Provost and Executive
Director at the California Digital Library, UC Office of the President.
As one of the world’s largest digital libraries, the California Digital
Library provides transformative digital library services, grounded in
campus partnerships and extended through external collaborations, that
amplify the impact of the libraries, scholarship, and resources of the
University of California. More at Günter
Waibel Staff Profile.
Nov 15: Cathy MARSHALL: What does it mean to turn every page in a
digital era?
In Working, journalist Robert Caro exhorts
researchers and journalists to "turn every page." But what does it
mean to turn every page in a digital era? There is little doubt that
the ready availability of online digital data has brought about a sea
change in research methods and results. Digital Humanities arose as a
field with the introduction of new digital corpora (some extensively
annotated) and more sophisticated analytic tools. Yet by now, the
category itself and its attendant disciplinary distinctions are blurring.
Other types of humanities researchers are taking increased advantage of
online resources as well: results from traditional foregrounded sources
like special collections and archives are being goosed by access to
background resources like structured databases (e.g. long-term weather
data), semi-structured text collections (e.g. large newspaper collections),
and creatively repurposed content from web services. Using examples from
fieldwork and my own ongoing collaborations, I will argue that these
emerging categories of online resources will have knock-on effects for
many aspects of research.
Nov 22: Final Progress Reports:
Michael WANG: Benchmarking Error and Robustness of AI
Text Detection Models.
As the capabilities of state-of-the-art LLMs become
more and more powerful, the challenge of detecting AI generated text
also becomes increasingly difficult. Research in this field has
yielded several promising methods, while also raising concerns about
error and robustness. A well-constructed benchmark can provide more
meaningful metrics to assess AI text detection models compared to
self-reported accuracy scores, which often overestimate the true
accuracy of AI text detection models.
Cora CAI: Maternal Nutrition and Child Health in the
Democratic Republic of Congo: Trends and Key Indicators (2007, 2013
and 2014).
The project focuses on data visualization to analyze
maternal nutrition and child health trends in the Democratic Republic
of Congo (2007, 2013, 2014). Key indicators such as iron and vitamin
A supplementation, child nutrition outcomes (stunting, anemia), and
maternal health factors (body mass, depression, mortality) are explored.
The analysis also includes socioeconomic factors like wealth index
and decision-making, as well as environmental factors like flood
events and food security.
Suzanne WONES, University Librarian, will be rescheduled
next semester.
Nov 29: No Seminar: Thanksgiving.
Dec 6: Last scheduled Seminar meeting: Clifford LYNCH.
The Seminar will resume in the spring semester.
Spring
2024 schedule and summaries.