Search Here ….

Tags

affect AI AI slop AI STORIES alumni art autocorrect Blogging book ChatGPT conferences data analysis digital art electronic literature ELMCIP environmental art exhibition flies Gephi history journalister LLMs London machine vision market literature motherhood network analysis norsk open access Pierre Huyghe publication publications research science fiction self-representation selfies social media Surveillance talks teaching tracking trip report twitter UiB visualisation
Uncategorized

Can human behaviour be predicted with enough data?

Jill3 Comments

I’m watching a presentation by Albert László Barabási on his book BURSTS: The Hidden Pattern Behind Everything We Do, and he’s telling some fascinating stories.

For instance, had you heard of Hasan Elahi, a media artist who after being investigated by the FBI for suspected terrorism (due to unusual travel patterns, among other things) put his entire life online at Tracking Transcience? Here’s a piece Elahi wrote in the New York Times about this project, or you can watch him talk about the project for TED. Actually the information is not very easy for a viewer to put together (he calls it user-unfriendly), but he has fed in a lot of images, his bank records, location, and so forth. The images are very sparse, void of people, factual. There are photos of every toilet he uses, for instance, or every taco he eats near a railway station, all time, location and date-stamped. But apparently, in BURST, Barabási analyses Elahi’s data to see how typical his patterns of movement are.

The reason for the title of BURSTS is that it turns out our behaviour is conglomerated in bursts (around 16 mins into the talk). We have periods where we send lots of emails, or talk on the phone a lot, or visit the library a lot, or even have sex a lot, and then there are gaps where we do these things much less. This actually follows power laws, and is not purely random.

Barabasi also analysed anonymised mobile phone data to analyse how people move around (around 24 mins into the video). He found that if you know the past movements of a person, you can predict their next location with a 93% accuracy. Another interesting point is that there is nobody whose predictability (in terms of location) is less than 80%. (That might actually explain why Foursquare and Gowalla and so on get a little boring after a while.) But ultimately, Barabási argues, with enough data, we might be able to predict all kinds of human behaviour – collapses of stock markets, wars, major historical events, and so on.

BURST has rather poor reviews on Amazon, where it is criticised for having too much story and too little science. Regardless of whether or not you read the book, his 30 minute talk on it is interesting.

Discover more from Jill Walker Rettberg

Subscribe to get the latest posts sent to your email.

3 thoughts on “Can human behaviour be predicted with enough data?”

  1. Jill Walker Rettberg

    I'm loving having two weeks before teaching starts. Lets me watch videos of people on our reading list talking. http://t.co/RgWRTFru

  2. Alessandro Galetto

    Can human behaviour be predicted with enough data? http://t.co/pYTfqjdI

  3. Inge Tang

    RT @TopsyRT: Can human behaviour be predicted with enough data? http://t.co/pmG72GAa

Leave A Comment

Recommended Posts

AI and algorithmic culture University politics

Academics in Norway: Sign this petition asking for research-based discussions of how to use AI in universities

I just signed a petition calling for Norwegian universities to use research expertise on AI when deciding how to implement it, rather than having decisions be made mostly administratively. ,  If you are a researcher in Norway, please read it and sign it if you agree – and share with anyone else who might be interested. The petition was written by three researchers at UiT: Maria Danielsen (a philosopher who completed her PhD in 2025 on AI and ethics, including discussions of art and working life), Knut Ørke (Norwegian as a second language), and Holger Pötzsch (a professor of media studies with many years of research on digital media, video games, disruption, and working life, among other topics).  This is not about preventing researchers from exploring AI methods in their research. It is about not uncritically accepting the hype that everyone must use AI everywhere without critical reflection. It is about not introducing Copilot as the default option in word processors, or training PhD candidates to believe they will fall behind if they do not use AI when writing articles, without proper academic discussion. Changes like these should be knowledge-based and discussed academically, not merely decided administratively, because they alter the epistemological foundations of research. Maria wrote to me a couple of months ago because she had read my opinion piece in Aftenposten in which I called for a strong brake on the use of language models in knowledge work. She was part of a committee tasked with developing UiT’s AI strategy and was concerned because there was so much hype and so few members of the committee with actual expertise in AI. I fully support the petition. There are probably some good uses for AI in research, but the uncritical, hype-driven insistence that we must simply adopt it everywhere is highly risky. There are many researchers in Norway with strong expertise in AI, language, ethics, working life, and culture. We must make use of this expertise. This is also partly about respect for research in the humanities, social sciences, psychology, and law. Introducing AI at universities and university colleges is not merely a technical issue, and perhaps not even primarily a technical one. It concerns much more: philosophy of science, methodological reflection, epistemology, writing, publishing, the working environment, and more. […]

screenshot of Grammarly - main text in the middle, names of experts on the left with reccomendations and on the right more info about the expert review feature
AI and algorithmic culture Teaching

Grammarly generated fake expert reviews “by” real scholars

Grammarly is a full on AI plagiarism machine now, generating text, citations (often irrelevant), “humanizing” the text to avoid AI checkers and so on. If you’re an author or scholar, they also have been impersonating and offering “feedback” in your name. Until yesterday, when they discontinued the Expert Review feature due to a class action lawsuit. Here are screenshots of how it worked.