Eran Toch

Eran Toch

@eranto.bsky.social

I invent new technologies and user experiences for privacy & human-AI interaction. Prof of Engineering at TAU.

494 followers96 following218 posts

Pinned

While we were in the shelter, I received an email announcing that my paper, co-authored with Bar Ifrah (whom I supervised on a master’s project) and @birnhack.bsky.social, has been accepted for publication in the Journal of Documentation >>

TheMarker just published its latest “40 Under 40” list. It got me thinking: are they actually that promising? I set Claude Code loose on a preliminary (and limited) analysis of whether the people featured in two lists over the years actually lived up to the promise. claude.ai/artifact/FZo...

View full thread

Interestingly, those selected also had about three times as many documented scandals, forced resignations, and major failures. The two lists are similarly predictive, although Globes seems to select somewhat riskier profiles.

We can’t tell whether the lists identified talented people or actually helped them succeed. Research on awards suggests that recognition itself can boost performance. There are plenty of other limitations, too. This is, after all, a fairly superficial analysis that Claude did in a few hours.

TheMarker just published its latest “40 Under 40” list. It got me thinking: are they actually that promising? I set Claude Code loose on a preliminary (and limited) analysis of whether the people featured in two lists over the years actually lived up to the promise. claude.ai/artifact/FZo...

Claude collected all the lists between 2006 and 2021: about 1,167 people. We traced their careers using open sources, mainly news archives, Wikipedia, and organizational websites. We also randomly selected 300 people and looked for comparable peers at a similar career stage. We found matches for 95.

The results surprised me. The lists did pretty well: 82% of those selected are still visible in the public sphere, and about a third reached a verifiable milestone—an exit, senior appointment, professorship, or board seat. Among matched peers, the difference was substantial: 31% vs. 13%.

Really happy to see my PhD student Vanessa Budde’s work recognized with a Short Paper Honorable Mention at MuC 2026 🎉. Our paper looks at how people experience harmful chatbot interactions: from manipulation and behavioral steering to boundary violations. toch.tau.ac.il/3924-2/

Congratulations to Michael Khavkin (my Ph.D. student) on receiving the Ein-Dor Student Research Award! The award, by ILAIS, was received for our paper “Beyond the Privacy-Utility Trade-Off: How User Disclosure Decisions Shape Information Systems Under Differential Privacy."

Claude Code is powerful, but as your projects pile up, it's hard to track everything across terminal sessions. Understudy is an open-source layer on top of Claude Code that makes it act less like a tool you drive and more like an assistant that keeps work moving on its own. 🧵 1/7

View full thread

Understudy can take real actions — sending messages, writing files — so safety is built in: a kill switch that freezes everything, optional per-project policies that limit what it may do, and every external action logged. Read SECURITY.md before connecting a real account. 6/7

security.mdHomepage

Claude Code is powerful, but as your projects pile up, it's hard to track everything across terminal sessions. Understudy is an open-source layer on top of Claude Code that makes it act less like a tool you drive and more like an assistant that keeps work moving on its own. 🧵 1/7

View full thread

State lives in readable Markdown and HTML on disk — no database, no lock-in. Read, diff, and back it up with tools you already have; it stays yours even if Understudy disappears, and you can spin any project off to its own Claude Code session. 4/7

Because Understudy keeps context across every project, it can surface what's in progress, flag projects that have stalled, and help you prioritize what actually deserves your attention. 5/7

Claude Code is powerful, but as your projects pile up, it's hard to track everything across terminal sessions. Understudy is an open-source layer on top of Claude Code that makes it act less like a tool you drive and more like an assistant that keeps work moving on its own. 🧵 1/7

It pairs a web dashboard with an autonomous queue. Each project is just a folder with an instruction. Understudy processes it in its own headless Claude Code session, then keeps the memory — status, history, decisions — in plain files between runs, so each run starts grounded. 2/7

The key: it runs unattended. Add work via the dashboard, Slack, or email, and the queue drains on a schedule while you're away — you come back to results waiting in the dashboard. Drop in a PDF noting "prepare an action plan for me," and next run, it's there. 3/7

Super proud of my PhD student Danielle Movsowitz Davidow, for giving her research, “The Blockchain Privacy Paradox: Perceptions of Privacy Coins,” at the 13th Privacy Conference at Tel Aviv University. >>

The lecture described interviews with cryptocurrency power users and sought to explain why privacy coins that offer strong cryptographic protections for user identities, such as Zcash and Monero, are disconnected from transactions with more traditional currencies.

336 units should suffice, at an estimated cost of 33m NIS. This estimate is based on a classic heuristic (Gonzalez k-center), under several assumptions: shelters have unlimited capacity (an assumption worth revisiting), differences in alert times are ignored, and no traffic congestion >>

Eran Toch reposted

Hashing isn’t a magic incantation that anonymizes your data.

M. Nolan Gray 🥑@mnolangray.bsky.social

What's something that experts/practitioners in your field universally agree upon, but that remains a "hot take" among the general public?

In the paper, "Temporal bias in historical newspaper digitisation — evidence from reconstructing the British Mandate trademark registry", we describe how we reconstructed a database of trademarks (i.e., logos) from the Palestine Gazette, the official newspaper of the British Mandate government >>

We developed a relatively simple learning algorithm that identifies trademark images and links them to textual information. Thanks to the newspaper’s fixed format, we were able to identify a systematic time-dependent bias, which can shape what we can see digitized sources.

Open-access publishing models allow well-funded academic institutions (such as my university) to use their financial resources and purchasing power to give their faculty an unfair advantage in the number of publications and citations. Isn't pay-to-publish more harmful than pay-to-read?

We’re happy to announce that our paper, “Place Matters: Neighborhood Effects on Smart City Service Adoption,” has been accepted for publication in the Journal of Urban Technology. >>

View full thread

In a new paper, we show that where you live matters as much as who you are. Using survey data from four very different neighborhoods in Tel Aviv, we find that neighborhood context explains technology adoption far better than demographics alone. >>

This matters because cities often design and promote these systems for certain residents and not others. User-facing digital services tend to cluster in certain neighborhoods, while city-led infrastructure, such as bike-sharing, shows a more equitable distribution.

We’re happy to announce that our paper, “Place Matters: Neighborhood Effects on Smart City Service Adoption,” has been accepted for publication in the Journal of Urban Technology. >>

Urban technologies, such as digital resident platforms like Tel Aviv’s DigiTel or parking apps, are now common in cities, yet adoption remains uneven despite major public investment. >>

Eran Toch reposted

“Let’s make this clear right now: unless you’ve made at least two masterpieces and changed the face of cinema forever — don’t walk on bike lanes.” Well played, Facebook account of the City of Tel Aviv-Jaffa

So proud of the work we’ve done in CHI’s Privacy and Security subcommittee. It was really a privilege co-chairing this subcommittee with Florian and Emilee. Sadly, our lab’s papers weren’t so lucky.

Eran Toch reposted

#CHI2026 program is shaping up: here's the first glimpse at the list of accepted workshops: chi2026.acm.org/workshops/ac... This will be a really fun @chi.acm.org ! (Reminder that there's no added fee this year to attend workshops, they will happen throughout the conference!)

chi2026.acm.orgAccepted Workshops - ACM CHI 2026CHI workshops are organized independently by their organizers. Please see the websites of the individual events for detailed instructions on how to submit position papers and/or applications. Please n...

Thank you, Alon Zoizner, for inviting me to speak at the Communication Department seminar at the University of Haifa and for the opportunity to present, for the first time, the work by Adam Aviv, Oshrat Ayalon, and me on Kosher phone users’ perceptions of safety, privacy, and community.

Eran Toch reposted

Eran Toch reposted

Final version! In Defense of Privacy-as-Control (Properly Understood), 65 Jurimetrics (2026 forthcoming) I defend Privacy-as-Control, explain why its wrong to reduce it to notice & choice, and argue it serves as a common denominator for privacy's many theories papers.ssrn.com/sol3/papers....

papers.ssrn.comIn Defense of Privacy-as-Control (Properly Understood)Privacy-as-control (‘PaC’) is one of the dominant conceptions of privacy. PaC means that each person should be able to decide for themselves whether to disclose

The main message is that our students are being hired into roles that require interdisciplinarity capabilities right from the start. We teach more Machine Learning and statistics than anyone else on campus, and our students train models themselves: the best way to engage with the world of AI.

We were awarded a very exciting new grant from the National Science Foundation (NSF) and the U.S.-Israel Binational Science Foundation (BSF) with colleagues Adam Aviv from The George Washington University and Oshrat Ayalon from the University of Haifa. >>

We will study the security and privacy concerns related to Haredi technologies, such as Kosher Phones. If you're interested in pursuing your Ph.D. or Master's degrees in this area, especially if you're a current or former member of Haredi communities, let us know!