Loved working with my amazing flat team of Fengli Xu and Lingfei Wu on our new PNAS paper "Flat Teams Drive Scientific Innovation" to show that relative to tall, hierarchical teams, flat, egalitarian ones produce more novelty and new directions for science
The amazing Johan Chu and I published a piece out this week at PNAS () showing how the growing size and success of sci/tech fields ironically impedes the rise of new ideas within them. A growing deluge of papers leads to a crystallization of the canon.
Thrilled to be hosting the International Conference on Computational Social Science (IC2S2) July 19-22, 2022 at UChicago with Sendhil Mullainathan, Rochelle Terman, Luis Bettencourt and a mind-blowing line-up of speakers: Details @ ; spread the word!
I'm hiring (several) postdocs
@KnowLab
this year and over the next 5 years at the University of Chicago in AI, Data and Computer Science, Computational Social Science, Tech Policy, Science of Science and Innovation, and more! Check out ads (and please forward to your favorite
I'm hiring a Senior Research Associate (PI eligible) at UChicago's Knowledge Lab to build community, programs, and catalyzing research at the intersection of knowledge (e.g., STS, innovation, [mis]information), AI, and society. Please apply and forward!
New piece out online today with Austin Kozlowski and Matt Taddy on the "Geometry of Culture" about how machine learning models built from historical text can identify cultural dimensions, trace cultural evolution, and enable cultural discovery: (1/2)
Finally, our book about 'Algorithmic Cultures' came out on the varied logics underlying AI in society and why need to understand, conserve, and cultivate algorithmic diversity. With Adrian Johns and many other rockstar historian and science studies friends...
New paper with
@arzhetsky
@valdanchev
demonstrates that centralized, networked scientific communities inadvertently produce much less replicable knowledge.
New paper in Nature Human Behaviour with the inimitable
@JamshidSourati
on how to build human-aware Artificial Intelligence (AI) that complements rather than substitutes for human capacity in creative, competitive domains like science and technology (S&T).
At Knowledge Lab I'm looking for a postdoc or predoc to start now & develop/apply computational methods to data on science and tech innovation for the broad challenge of tracking the direction of global advance and what governments should fund! Pls share!
Come join us for an amazing line-up at the International Conference on the Science of Science and Innovation (ICSSI) at the National Academies in Washington D.C. July 1-3, 2024! Submissions by March 24!
Here is our take what is computational sociology...and computational social science). Our answer suggests that computation should not only shape how we answer social questions, but how expand our imagination to ask them (w UCLA's Jacob Foster):
Excited about my recent paper led by the amazing Feng (Bill) Shi in Nature Communications on How Science and Technology are Driven by Surprise, as driven by Expeditions from Intellectual & Technical Outsiders:
With the amazing Khwan Kim and Noah Askin, we just published a paper in PNAS on how tastes for culture and music evolve worldwide, suggesting that we (all) construct the playlists of our lives.
Let's be friends ;-). I'm hiring an Instructional Professor in Computational Social Science in the Master of Arts Program in Computational Social Science (MACSS) at the University of Chicago!
June 7-9, 2022 I'm thrilled to be co-organizing the International Conference on the Science of Science (ICSSI) with the AMAZING Heidi Williams &
@dashunwang
at the National Academies in DC. Come present your most awesome work; submit through 3/22! (@ )
How does language shape our ideas, conversations, and collaborations? New work with the amazing Pedro ('Pete') Aceves
@JHUCarey
discovers how human languages vary widely in how densely they encode meaning and its striking consequences for communication and collaboration!
Roots! I "entered" social science as an anthropology major 35 years ago with an interest in how meanings were organized across cultures, and finally published a paper that provides new insight led by the amazing
@mollyllewis
in PNAS.
Preprint:
In a world of intense competition, failure is a prerequisite for success. Our latest paper in Nature out today reveals that across science, start-ups and security domains, accelerating failures strongly foreshadow upcoming success: (1/6)
#innovation
What kind of AI do we deserve?
My TEDx talk () builds an argument for expanding human capacity and imagination not by building the most, but the LEAST human AIs -- beneficial Alien Intelligences designed to complement our individual and collective
Flat teams increase productivity for those on bottom and increase long-term impact, suggesting the critical importance of inviting and involving more "brains" to the table for generating sustainable advance:
Cover art for my new book, Craft & Code, with Adrian Johns and a crew of awesome historian friends...drawn by Christian Clarke & Dalle-2 "prompted" by me and the text itself.
🤔What's the role of co-location in scientific discovery? The usual measures underestimate it!
🚨New paper by Eamon Duede
@phil_of_ai
(with
@profjamesevans
@klakhani
and me)
Open access:
(1/3)
Delighted to share amazing work
#FAccT2023
led by the amazing
@NandanaSengupta
on how biased data has real consequences: In AI models of urban safety, more data from male evaluators DECREASE the accuracy of AI predictions for women's experience of safety!
Featured article out today in the New Yorker about a class on existential risks taught at UChicago by the inimitable Daniel Holz (
@decohere
) and myself this spring:
Come be my friend! Check out a new opportunity teaching computational social science to amazing students in a vibrant intellectual community at the University of Chicago!
Review just out with the amazing
@LikunCao
and
@ZiwnChen
on the paradox of "innovation science"--when a pattern becomes predictable, it ceases to be innovation. The concept of 'creative destruction' has focused on how what emerges wipes out what came before; but...
Mindblowing first day at the International Conference on Science of Science and Innovation
#ICSSI2023
@NorthwesternU
!with panels on AI and science ; gender and science, and a host of amazing presentations on the mechanics and impact of science in society!
As part of a project I run focused on improving the explanatory and prescriptive power of social science models (see ), I have engaged a crowd-sourcing firm, , to launch a series of analysis challenges over the coming months (1/6)
Had a mind-bending blast doing the Complexity podcast with Santa Fe Institute’s Michael Garfield this week on Social Computing and Designing Diversity with AI:
Social relationships have been hard hit the last few weeks--they facilitate infection and spread disease. In a recent paper with amazing colleagues Linzhuo Li and
@LingfeiWu
, we kick them while they're down (sorry fellow sociologists)
New work out with the inimitable Jacob Foster (IU) in Critical Inquiry about how to build "bad robots"--creative artificial intelligence designed to provoke and augment humanistic creativity rather than mimic it. ("Algorithmic Abduction: Robots for Alien Reading":
New work out with the inimitable
@ishanu_ch
and colleagues that uses big urban crime data and novel models to surveil the surveillers--uncovering biased enforcement that draws policing resources to privileged neighborhoods when crime spikes in U.S. cities:
Amazing speakers @ include Audrey Tang, Rediet Abibe, Chris Bail, Tom Griffiths, Alex Chouldechova, Iyad Rahwan, Lillian Lee, Christian Danescu Niculescu Mizil, Molly Roberts, Ted Underwood & more.
Glad to share a new paper
@PNASNexus
where we use BERT on millions of docs to show that prescient ideas come from the periphery in business, law and politics.
Here is how we do it and what we find >>
Co-authored w/ Paul Vicinanza & Sameer Srivastava, Computational Culture Lab
We're in a battle between traditionalism and novelty.
@profjamesevans
' from
@UChicago
approaches how we might stimulate more valid and surprising outcomes through diversity and AI.
Hear more at his seminar next Monday 5 pm CEST!
We highlight data-driven and computation opportunities to understand innovation in ways suggested by complexity, science studies, and science fiction, and to manage it. Here is the paper (); and for free ()
We show that increased social connection and centralization can also be associated with decreased diversity. Because diversity drives innovation, this creates an essential tension between the demand for and supply of difference.
I enjoyed a delightfully deep and far-ranging podcast discussion with Ryan; thanks for having me on Origins! Check out the chat and/or my accompanying (mostly musical) playlist ;-)!
New paper in
@PhysRevLett
! We use our Bayesian machine scientist to compare data collapses rigorously, and shed light upon a 80+ year old problem in turbulence. Great work by
@turboignasi
Finally, we evaluated the likelihood that each claim will replicate by matching them to the results of a massively replicated high-throughput experiment.
The opportunity? Science policies that embrace failure and facilitate independent replication through greater competition and diversity will generate more robust knowledge that can make a difference for human health and prosperity.
The risk is that this sets up its own bias: Any out-of-the-box idea that’s wrong is going to be uninteresting to most of the world. The ideas where any outcome is “interesting” are promotions or negations of phenomena everyone already knows and cares about.
@NAChristakis
As someone in said society…this just makes me sad. Qualitative sociology could/should also be reproducible…or it becomes fine art, an irreducible convolution of subject and object.
Markov chains & processes were invented to show that dependencies between outcomes could still obey the central limit theorem, countering Nekrasov’s argument that because Russian crime reports obeyed the law of large numbers, crime reflected acts of free will.
We illustrate this with the meanings of class over the 20th Century, but with implications ranging from high dimensional theorizing to applications in search engine optimization and recommender systems where context sensitivity matters. (2/2)
New article out today with
@NandanaSengupta
and Nati Srebro validating "Simple Surveys"--incomplete, computationally customized questionnaires--currently used to drive recommendations on social and digital media platforms, for better social science:
We also discover that (2) concrete meanings are less variable across languages than abstract ones, but all vary with geographical and cultural (e.g., kinship, law) distance.
The alignment of these meaning spaces predictively reveals which metaphors, analogies, and associations within any one language translate or confuse in every other language. Summary: Paywall:
Come be my friend! Check out a new opportunity teaching computational social science to amazing students in a vibrant intellectual community at the University of Chicago!
For math & stats nerds, amazing how many innovations have emerged from critical studies of criminal justice--Poisson's distribution was discovered analyzing the number of wrongful convictions in a country;
Send your great papers and tutorials (extended deadline THIS FRIDAY, 3/11) for the International Conference on Computational Social Science (IC2S2) July 19-22, 2022 at UChicago () New grad student lodging and scholarship options available!
The details? We analyzed tens of thousands of published biomedical claims about drug-gene interactions, identifying the networks of scientists, methods and knowledge that produced them.
We coin 'destructive creation' to focus the vast, fragmented research on how innovation emerges from discord and disorder, at large scales and small. We highlight new data-driven opportunities to understand and surf this critical boundary between order and chaos...
Our precise predictions (90% accuracy for property and violent crimes predicted 7 days in advance within ~1,000 ft) are based on a novel machine learning model inspired by the depth of neural networks but with a very different underlying architecture tuned to events in time.
A recipe for research with no risk of incentive to hide the truth may also, unintentionally, be a recipe for riskless (boring) research. [Which, to be clear, is the opposite of my impression of YOUR research! I’m a huge fan.]
Whether you want to foment success or thwart it, our findings unveil strong early signals that identify whether failure dynamics will lead to ultimate success or failure. (6/6)
Detailed summary @
#Innovation
#ScienceOfScience
Exciting new work on the growing role of federally funded research in patenting (which is partly a story of the eclipse of small inventors by big science/tech).
Should governments invest in basic research? Out today in
@sciencemagazine
: we report that nearly one-third of patented U.S. inventions rely on federally-funded science, a rate which has grown substantially over the past 90 years:
(with Tyler Reigeluth, Stephanie Dick, Clare Kim, Alex Csiszar, Honghong Tinn, Salem Elzway, Matthew Jones, Alma Steingart, Mike Annany, Hallam Stevens, Theodora Dryer, Ksenia Tatarchenko, John Tresch)!
Our project aims to use social science insights, statistical analysis, and reoriented computational models to emphasize explanation, counterfactual prediction and intervention. (6/6)
🔥Try this Chrome plugin!🔥 It lets you use Twitter to learn about new science/papers etc - while avoiding most of the toxic, unpleasant content that can make this place a hellscape. This has saved Twitter for me! Hats off to
@hauselin
292 people from 11 locations around the world participated in the Summer Institutes in Computational Social Science this year, creating countless new research teams, 4,210 Github Commits, and one very lonely word cloud. Your 2019 slideshow,
#SICSS2019
:
New chapter of ours out that shows how ideas and inventions—like Zaphod from the Hitchhiker's Guide to the Galaxy or bacteria with horizontal gene transfer—typically have many mothers, and how to overhaul the modeling to better understand and predict them!
With more than one hundred million musical streams by tens of thousands of international listeners, we show that despite speculation about how musical nostalgia compensates for life’s disruptions, musical exploration systematically reflects breaches in personal routine.
Those who eventually succeed and those who do not initially appear similar, but are characterized by fundamentally distinct failure dynamics in the efficiency and quality of each attempt. (3/6)
#Innovation
#KnowledgeSquared
#ScienceOfScience
Our analysis is based on the very precise prediction of recorded crime events, recognizing that such events represent an interaction between reported (mis)behavior, surveillance, and trust in policing, summarized here:
Our results reveal a hidden source of variation across languages and reveal how it shapes the nature and texture of human engagement, with consequences for human collaboration and conflict across society. ; free:
BUT the same semantically dense languages produce conceptually narrower conversations and expositions, where topics are recycled and ultimately discussed in substantially greater depth?!
We evaluate our model among entrepreneurs exiting their startups, scientists seeking to fund their research, and terrorist groups trying to post casualties, finding broadly consistent patterns. (5/6)
#Innovation
#KnowledgeSquared
#ScienceOfScience
@SarahSbratt
Starting the band...now! I'm on trombone (always an accident). Yes, moving to new institutions does moderate the influence, but it takes time for these within institution / not in department connections to yield.