Jessica Hullman
@jessicahullman.bsky.social
Ginni Rometty Prof @NorthwesternCS | Fellow @NU_IPR | Uncertainty + decisions | Humans + AI/ML | Blog @statmodeling
Pinned
ExplainableAI has long frustrated me by lacking a clear theory of what an explanation should do. Improve use of a model for what? How? Given a task what's max effect explanation could have? It's complicated bc most methods are functions of features & prediction but not true state being predicted 1/
Reposted by Jessica Hullman
If you work at the intersection of CS and economics (or think your work is of interest to those who do!) consider submitting to the ESIF Economics and AI+ML meeting this summer at Cornell: www.econometricsociety.org/regional-act...
2026 ESIF Economics and AI+ML Meeting - The Econometric Society
2026 ESIF Economics and AI+ML Meeting (ESIF-AIML2026) June 16-17, 2026 Cornell University Department...
www.econometricsociety.org
November 8, 2025 at 10:37 PM
If you work at the intersection of CS and economics (or think your work is of interest to those who do!) consider submitting to the ESIF Economics and AI+ML meeting this summer at Cornell: www.econometricsociety.org/regional-act...
ICE showed up to kidnap lawn workers and nannies today in my sleepy Chicago North Shore neighborhood. So disgusted
November 8, 2025 at 12:17 AM
ICE showed up to kidnap lawn workers and nannies today in my sleepy Chicago North Shore neighborhood. So disgusted
Reposted by Jessica Hullman
I’m recruiting students this upcoming cycle at UIUC! I’m excited about Qs on societal impact of AI, especially human-AI collaboration, multi-agent interactions, incentives in data sharing, and AI policy/regulation (all from both a theoretical and applied lens). Apply through CS & select my name!
November 6, 2025 at 6:52 PM
I’m recruiting students this upcoming cycle at UIUC! I’m excited about Qs on societal impact of AI, especially human-AI collaboration, multi-agent interactions, incentives in data sharing, and AI policy/regulation (all from both a theoretical and applied lens). Apply through CS & select my name!
Reposted by Jessica Hullman
Slides from my @mit.edu IDSS Distinguished Speaker seminar "Networks untangle gender differences in productivity and prominence among scientists" this week
I argue that collaboration networks act like unequally distributed (and gendered) social capital
aaronclauset.github.io/slides/Claus...
I argue that collaboration networks act like unequally distributed (and gendered) social capital
aaronclauset.github.io/slides/Claus...
November 6, 2025 at 6:05 PM
Slides from my @mit.edu IDSS Distinguished Speaker seminar "Networks untangle gender differences in productivity and prominence among scientists" this week
I argue that collaboration networks act like unequally distributed (and gendered) social capital
aaronclauset.github.io/slides/Claus...
I argue that collaboration networks act like unequally distributed (and gendered) social capital
aaronclauset.github.io/slides/Claus...
🧠⚙️ Interested in decision theory+cogsci meets AI? Want to create methods for rigorously designing & evaluating human-AI workflows?
I'm recruiting PhDs to work on:
🎯 Stat foundations of multi-agent collaboration
🌫️ Model uncertainty & meta-cognition
🔎 Interpretability
💬 LLMs in behavioral science
I'm recruiting PhDs to work on:
🎯 Stat foundations of multi-agent collaboration
🌫️ Model uncertainty & meta-cognition
🔎 Interpretability
💬 LLMs in behavioral science
November 5, 2025 at 4:40 PM
🧠⚙️ Interested in decision theory+cogsci meets AI? Want to create methods for rigorously designing & evaluating human-AI workflows?
I'm recruiting PhDs to work on:
🎯 Stat foundations of multi-agent collaboration
🌫️ Model uncertainty & meta-cognition
🔎 Interpretability
💬 LLMs in behavioral science
I'm recruiting PhDs to work on:
🎯 Stat foundations of multi-agent collaboration
🌫️ Model uncertainty & meta-cognition
🔎 Interpretability
💬 LLMs in behavioral science
A very reasonable take on AI in education
Wrote a short piece arguing that higher ed must help steer AI. TLDR: If we outsource this to tech, we outsource our whole business. But rejectionism is basically stalling. If we want to survive, schools themselves must proactively shape AI for education & research. [1/6, unpaywalled at 5/6] +
Opinion | AI Is the Future. Higher Ed Should Shape It.
If we want to stay at the forefront of knowledge production, we must fit technology to our needs.
www.chronicle.com
November 4, 2025 at 8:02 PM
A very reasonable take on AI in education
Tonight’s mood…
No offense to John Denver but this version captures something the original missed
No offense to John Denver but this version captures something the original missed
The Eagle And The Hawk
YouTube video by Bonnie Prince Billy - Topic
youtu.be
November 4, 2025 at 1:52 AM
Tonight’s mood…
No offense to John Denver but this version captures something the original missed
No offense to John Denver but this version captures something the original missed
Reposted by Jessica Hullman
Two answers:
- Anthropomorphization makes sense when dealing with written human-like characters, which is what LLMs generate
- We aren’t very deep into interpretability yet
x.com/pfau/status/...
- Anthropomorphization makes sense when dealing with written human-like characters, which is what LLMs generate
- We aren’t very deep into interpretability yet
x.com/pfau/status/...
October 31, 2025 at 3:53 PM
Two answers:
- Anthropomorphization makes sense when dealing with written human-like characters, which is what LLMs generate
- We aren’t very deep into interpretability yet
x.com/pfau/status/...
- Anthropomorphization makes sense when dealing with written human-like characters, which is what LLMs generate
- We aren’t very deep into interpretability yet
x.com/pfau/status/...
Reposted by Jessica Hullman
When referring to an unspecified, imaginary person, why do some writers need to refer to them specifically as “she” or “he”?
🧠: are you proudly showing that your imagination is this specific or are you just really against using “they”?
Either way, it’s a distraction.
🧠: are you proudly showing that your imagination is this specific or are you just really against using “they”?
Either way, it’s a distraction.
October 30, 2025 at 8:36 AM
When referring to an unspecified, imaginary person, why do some writers need to refer to them specifically as “she” or “he”?
🧠: are you proudly showing that your imagination is this specific or are you just really against using “they”?
Either way, it’s a distraction.
🧠: are you proudly showing that your imagination is this specific or are you just really against using “they”?
Either way, it’s a distraction.
Great post on various ways in which multiverse analysis can be misdirected as a solution
People seem to be discussing multiverse analysis again! I haven’t read the latest piece but will use that as an opportunity to share a blog post of mine with a title of which I’m still proud (although I probably shouldn’t)
www.the100.ci/2021/03/07/m...
www.the100.ci/2021/03/07/m...
Mülltiverse Analysis
Psychologists like their analyses like I like my coffee: robusta. Results shouldn’t change too much, no matter which exclusion criteria are applied, which covariates are included, which transformation...
www.the100.ci
October 26, 2025 at 4:49 PM
Great post on various ways in which multiverse analysis can be misdirected as a solution
✨ SOTA in Probability Elicitation, 2025 ✨
Econ/stats/JDM: "We invoke Schervish (1989) to derive a task-aligned strictly proper scoring rule, and expose subjects to 50 calibration trials to verify comprehension. We provide proofs of LIC in ..."
HCI: “We used a slider to promote carefulness.”
Econ/stats/JDM: "We invoke Schervish (1989) to derive a task-aligned strictly proper scoring rule, and expose subjects to 50 calibration trials to verify comprehension. We provide proofs of LIC in ..."
HCI: “We used a slider to promote carefulness.”
October 25, 2025 at 1:47 AM
✨ SOTA in Probability Elicitation, 2025 ✨
Econ/stats/JDM: "We invoke Schervish (1989) to derive a task-aligned strictly proper scoring rule, and expose subjects to 50 calibration trials to verify comprehension. We provide proofs of LIC in ..."
HCI: “We used a slider to promote carefulness.”
Econ/stats/JDM: "We invoke Schervish (1989) to derive a task-aligned strictly proper scoring rule, and expose subjects to 50 calibration trials to verify comprehension. We provide proofs of LIC in ..."
HCI: “We used a slider to promote carefulness.”
Why are there so many houses with bathrooms in basements? Especially bathrooms with showers in unfinished basements. Why on earth? Why does my own house have a bathroom in the basement? Who is going to use it?
October 24, 2025 at 1:57 AM
Why are there so many houses with bathrooms in basements? Especially bathrooms with showers in unfinished basements. Why on earth? Why does my own house have a bathroom in the basement? Who is going to use it?
I’m giving an IDE seminar at @mitsloan.bsky.social tomorrow at 11am, on optimizing AI as decision support. Joint work w/ @ziyang.bsky.social @yifanwu.bsky.social @jasonhartline.bsky.social @berkustun.bsky.social
Come by if you’re around!
www.eventbrite.com/e/fall-2025-...
Come by if you’re around!
www.eventbrite.com/e/fall-2025-...
October 22, 2025 at 7:06 PM
I’m giving an IDE seminar at @mitsloan.bsky.social tomorrow at 11am, on optimizing AI as decision support. Joint work w/ @ziyang.bsky.social @yifanwu.bsky.social @jasonhartline.bsky.social @berkustun.bsky.social
Come by if you’re around!
www.eventbrite.com/e/fall-2025-...
Come by if you’re around!
www.eventbrite.com/e/fall-2025-...
Reposted by Jessica Hullman
And now might be a good time to mention, I’m on the faculty job market this year! I do work in human-AI collusion, collaboration and competition, with an eye towards building foundations for trustworthy AI. Check out more info on my website here! Nataliecollina.com
October 22, 2025 at 3:56 PM
And now might be a good time to mention, I’m on the faculty job market this year! I do work in human-AI collusion, collaboration and competition, with an eye towards building foundations for trustworthy AI. Check out more info on my website here! Nataliecollina.com
@abhsarma.bsky.social created a visualization tool for principled reasoning about whether analysis paths should be included in a multiverse
Try it: abhsarma.github.io/milliways/
Paper:
mucollective.northwestern.edu/project/2024...
Code:
github.com/abhsarma/mil...
Try it: abhsarma.github.io/milliways/
Paper:
mucollective.northwestern.edu/project/2024...
Code:
github.com/abhsarma/mil...
October 21, 2025 at 4:56 PM
@abhsarma.bsky.social created a visualization tool for principled reasoning about whether analysis paths should be included in a multiverse
Try it: abhsarma.github.io/milliways/
Paper:
mucollective.northwestern.edu/project/2024...
Code:
github.com/abhsarma/mil...
Try it: abhsarma.github.io/milliways/
Paper:
mucollective.northwestern.edu/project/2024...
Code:
github.com/abhsarma/mil...
So excited
No one tells you that full professor really means full return to your high school self, but that’s basically what it is
No one tells you that full professor really means full return to your high school self, but that’s basically what it is
October 19, 2025 at 1:09 AM
So excited
No one tells you that full professor really means full return to your high school self, but that’s basically what it is
No one tells you that full professor really means full return to your high school self, but that’s basically what it is
Reposted by Jessica Hullman
"Wank is a speech act directed first and foremost at helping the speaker feel better about themselves, stated as an objective claim about the state of the world, that we are expected in discussion to treat with the epistemic authority of a claim about identity, but the content of the actual claim".
The various people trying to shame me for using large language models to do actual research and produce detailed teaching materials on their harms, instead of simply preaching against them on a street corner—they provide a perfect illustration of wank.
An essay on wank | deadSimpleTech
This captures well the uncomfortable, slightly disorienting feeling that wank creates when you're subjected to it, wherein you're expected to speak about and think about the statement as though it say...
deadsimpletech.com
October 18, 2025 at 3:50 PM
"Wank is a speech act directed first and foremost at helping the speaker feel better about themselves, stated as an objective claim about the state of the world, that we are expected in discussion to treat with the epistemic authority of a claim about identity, but the content of the actual claim".
Reposted by Jessica Hullman
Since I'm about to head out to DC's No Kings protest, here's a quick look at America's ORIGINAL No Kings movement, and a review of how many of George III's crimes outlined in the Declaration have been repeated by Trump. juliansanchez.substack.com/p/trump-vs-g...
Trump vs. George III: A No Kings Comparison
Our own mad king is repeating many of his predecessors' greatest hits
juliansanchez.substack.com
October 18, 2025 at 3:21 PM
Since I'm about to head out to DC's No Kings protest, here's a quick look at America's ORIGINAL No Kings movement, and a review of how many of George III's crimes outlined in the Declaration have been repeated by Trump. juliansanchez.substack.com/p/trump-vs-g...
Reposted by Jessica Hullman
It’s become fashionable in some circles to reject decision theory (and other basic statistical ideas) for vaguely political reasons.
There are valid critiques—but also some worth being wary of.
Some thoughts:
statmodeling.stat.columbia.edu/2025/10/17/s...
There are valid critiques—but also some worth being wary of.
Some thoughts:
statmodeling.stat.columbia.edu/2025/10/17/s...
Separating the whack from the chaff in critiques of decision theory | Statistical Modeling, Causal Inference, and Social Science
statmodeling.stat.columbia.edu
October 17, 2025 at 4:21 PM
It’s become fashionable in some circles to reject decision theory (and other basic statistical ideas) for vaguely political reasons.
There are valid critiques—but also some worth being wary of.
Some thoughts:
statmodeling.stat.columbia.edu/2025/10/17/s...
There are valid critiques—but also some worth being wary of.
Some thoughts:
statmodeling.stat.columbia.edu/2025/10/17/s...
It’s become fashionable in some circles to reject decision theory (and other basic statistical ideas) for vaguely political reasons.
There are valid critiques—but also some worth being wary of.
Some thoughts:
statmodeling.stat.columbia.edu/2025/10/17/s...
There are valid critiques—but also some worth being wary of.
Some thoughts:
statmodeling.stat.columbia.edu/2025/10/17/s...
Separating the whack from the chaff in critiques of decision theory | Statistical Modeling, Causal Inference, and Social Science
statmodeling.stat.columbia.edu
October 17, 2025 at 4:21 PM
It’s become fashionable in some circles to reject decision theory (and other basic statistical ideas) for vaguely political reasons.
There are valid critiques—but also some worth being wary of.
Some thoughts:
statmodeling.stat.columbia.edu/2025/10/17/s...
There are valid critiques—but also some worth being wary of.
Some thoughts:
statmodeling.stat.columbia.edu/2025/10/17/s...
Reposted by Jessica Hullman
A Brief History of Let’s Bring Back The Good Old Days
🧵
🧵
October 15, 2025 at 9:54 PM
A Brief History of Let’s Bring Back The Good Old Days
🧵
🧵
Think of how much better off we'd be if every established researcher got in the habit of writing papers entitled "Second thoughts on [thing I'm famous for]"
October 11, 2025 at 3:49 AM
Think of how much better off we'd be if every established researcher got in the habit of writing papers entitled "Second thoughts on [thing I'm famous for]"
Reposted by Jessica Hullman
A reoccurring frustration for philosophers of science: Many scientists know how to do science like people know how to ride a bike. When they reflect on the practice of science, they repeat platitudes about how science works. Those platitudes are often wrong, sometimes even about their own field
*sighs in philosopher of science*
Looking for confirmatory evidence is an entirely normal part of science. The primary problem here is the eugenics and the fascism, not the lies to children about "the scientific method."
Looking for confirmatory evidence is an entirely normal part of science. The primary problem here is the eugenics and the fascism, not the lies to children about "the scientific method."
October 10, 2025 at 3:26 PM
A reoccurring frustration for philosophers of science: Many scientists know how to do science like people know how to ride a bike. When they reflect on the practice of science, they repeat platitudes about how science works. Those platitudes are often wrong, sometimes even about their own field
Reposted by Jessica Hullman
Anyway I wrote an article back in 2013 analyzing permutations on Benedict Cumberbatch which you may read and sub in Timothée Chalamet should you so desire
web.archive.org/web/20250322...
web.archive.org/web/20250322...
A Linguist Explains the Rules of Summoning Benedict Cumberbatch -The Toast
A willing foe, and sea room.
web.archive.org
October 9, 2025 at 9:25 PM
Anyway I wrote an article back in 2013 analyzing permutations on Benedict Cumberbatch which you may read and sub in Timothée Chalamet should you so desire
web.archive.org/web/20250322...
web.archive.org/web/20250322...
I'm giving keynote at a workshop on predictive uncertainty for decisions tomorrow & realizing I may have more questions than answers
What do ppl really do w/conformal prediction sets?
What should they do?
What to do when calibration isn't enough for trust?
What's shape of uncertainty in high dim?
What do ppl really do w/conformal prediction sets?
What should they do?
What to do when calibration isn't enough for trust?
What's shape of uncertainty in high dim?
October 9, 2025 at 3:48 PM
I'm giving keynote at a workshop on predictive uncertainty for decisions tomorrow & realizing I may have more questions than answers
What do ppl really do w/conformal prediction sets?
What should they do?
What to do when calibration isn't enough for trust?
What's shape of uncertainty in high dim?
What do ppl really do w/conformal prediction sets?
What should they do?
What to do when calibration isn't enough for trust?
What's shape of uncertainty in high dim?