Remote

Michael Bernstein

@msbernst@hci.social

mastodon 4.7.3

Stanford CS prof, human-computer interaction and social computing. Surely I will be interesting and entertaining here, even though I failed to achieve this on Twitter.

933 Followers

227 Following

27 Posts

Joined May 03, 2022

Michael Bernstein @msbernst@hci.social

· 37mo ago

I've been wrestling with the gap between HCI curricula, which tend to focus on the design process, and the big theories and ideas that animate us in HCI. Here's a model that has starting working well here for me here at Stanford: @msbernst@medium.com

32

4

16

0

Michael Bernstein @msbernst@hci.social

· 21mo ago

Big congratulations to Helena Vasconcelos on winning a 2025 CRA Outstanding Undergraduate Researcher Award! https://cra.org/about/awards/outstanding-undergraduate-researcher-award/

8

0

0

0

Michael Bernstein @msbernst@hci.social

· 23mo ago

Coming very soon: #cscw2024 panel on "Is Human-AI Interaction CSCW?" alongside @asb@hci.social, @jbigham@hci.social, @andresmh@hci.social, and Merrie Morris!

https://programs.sigchi.org/cscw/2024/program/content/179232

🌱 hci.social

5

0

5

0

Michael Bernstein @msbernst@hci.social

· 31mo ago

8

0

3

0

Michael Bernstein @msbernst@hci.social

· 46mo ago

Replying to

I think that this leaves us with a challenge: we don't have many explainability techniques that are substantially less work to verify than completing the task manually.

15

1

1

0

Michael Bernstein @msbernst@hci.social

· 23mo ago

Replying to

A reviewer called @lindsay's work "a rare example of a 'full stack' CSCW paper. It is an unusual system paper that feels theoretically motivated and is evaluated experimentally against a meaningful control." Congratulations, @lindsay!

4

1

0

0

Michael Bernstein @msbernst@hci.social

· 21mo ago

Stanford's Symbolic Systems Program is hiring a three-year lecturer position: https://academicjobsonline.org/ajo/program/29530

academicjobsonline.org

3

0

3

0

Michael Bernstein @msbernst@hci.social

· 46mo ago

Replying to

Prior work has only been able to reduce overreliance by forcing people to engage critically with the AI's explanation. Helena Vasconcelos's new #cscw2023 paper finally---finally!---demonstrates that explanations _can_ reduce overreliance and increase decision-making performance even without forcing functions: https://arxiv.org/abs/2212.06823

10

1

0

0

Michael Bernstein @msbernst@hci.social

· 46mo ago

Replying to

Our hunch was that many of the null results fell into a valley where the cognitive cost of verifying the AI's explanation were roughly the same as the cognitive costs of just doing the task yourself. When doing it yourself and verifying the AI are both costly, people ignore the explanations and just rely on the AI.

9

1

0

0

Michael Bernstein @msbernst@hci.social

· 31mo ago

You just got asked to review another paper; your third emoji is your review.

4

6

2

0

Michael Bernstein @msbernst@hci.social

· 31mo ago

Replying to

Our argument is that design patterns largely cluster within cells of the Form-From design space. For example, threaded spaces commonly utilize upvotes, threaded networks often use resharing, and flat networks typically require posting on a "wall" rather than a central feed. If you don't like our 2x3 space, there's also a full 62-dimension treatment in the appendix from our inductive process, as well as a set of 11 categories those fall into. They're more useful for fine-grained distinctions.

3

1

0

0

Michael Bernstein @msbernst@hci.social

· 46mo ago

Replying to

Helena demonstrated through a series of five studies that if we manipulate the difficulty of the task and explanation, people do a cost-benefit tradeoff: they reduce overreliance and increase accuracy when explanations save significant effort.

4

1

0

0

Michael Bernstein @msbernst@hci.social

· 31mo ago

@andresmh@hci.social Yes, the main difference is that the network is at best secondary and possibly just nonexistent. For example, you can follow on TikTok, but it's not a main driver of the FYP.

In theory, you can algorithmically rank any threaded design: spaces (Reddit), network (FB/Insta feed), or commons (TikTok FYP). Commons-based designs have few options beyond algorithmic ranking, though.

And yeah, when there's a bounded space, paid subscriptions to gain access is a clear metaphor.

2

0

0

0

Michael Bernstein @msbernst@hci.social

· 46mo ago

Replying to

@Helena led this work, with collaborators Matthew Jörke, Madeleine Grunde-McLaughlin, @tobigerstenberg , me, and Ranjay Krishna. Go Helena!

3

0

0

0

Michael Bernstein @msbernst@hci.social

· 36mo ago

Replying to

@Niloufar Love this! And sorry I missed it, something was off with my Toot notifications! I should do this with my social computing final projects.

1

0

0

0

Michael Bernstein @msbernst@hci.social

· 37mo ago

Replying to

@jordanwb @evanmpeck Cool! Yes, we wound up splitting off our advanced design studio course from this more theory-driven course, since they are such different genres of class.

1

0

0

0

Michael Bernstein @msbernst@hci.social

· 31mo ago

Replying to

One of my favorite little investigations was coding all the systems listed in a Wikipedia social media timeline (https://en.wikipedia.org/wiki/Timeline_of_social_media) as of when they launched with Form-From. The pattern weaves back and forth between flat systems and threaded systems over time.

en.wikipedia.org

0

1

0

0

Michael Bernstein @msbernst@hci.social

· 33mo ago

Replying to

We're looking for a broad set of scholars, ranging across qualitative and critical studies, quantitative data science or experimental research, policy, social computing platform design and deployment, and AI. Apply by February 20.

0

0

0

0

Michael Bernstein @msbernst@hci.social

· 31mo ago

Replying to

So just what is this thing? It's been a long time since CSCW introduced Johansen's Time-Space matrix, and at this point a vast majority of social media would fall into the different time - different place quadrant, making it not very productive as a design or theory tool.

0

4

0

0

Michael Bernstein @msbernst@hci.social

· 31mo ago

Replying to

Form-From asks two questions: (1) What is the principal shape, or form, of the content: either threaded or flat? (2)~From where or from whom one might receive content, ranging from spaces to networks to the commons?

0

3

0

0

Michael Bernstein @msbernst@hci.social

· 31mo ago

Replying to

@wollman @axz @karger Yup, it's in there! Though would love to hear any additional color you can add to the story.

0

1

0

0

Michael Bernstein @msbernst@hci.social

· 46mo ago

Replying to

@josephseering Woohoo!!!

0

1

0

0

Michael Bernstein @msbernst@hci.social

· 37mo ago

Replying to

@evanmpeck Yeah, I think it's also hard to teach theory at the same time as you're teaching the intro to the process. Going too meta too fast gets confusing.

0

0

0

0