Wim De Neys
@wimdeneys.bsky.social
Psychological scientist @ CNRS & University of Paris
Reposted by Wim De Neys
10/10
Huge thanks to my great co-authors @zoepurcell.bsky.social , @luciecharlesneuro.bsky.social and @wimdeneys.bsky.social, and to my lab @lapsyde.bsky.social.
Stay tuned for the computational modeling part! 🤓
You can access the preprint here: osf.io/preprints/ps...
Huge thanks to my great co-authors @zoepurcell.bsky.social , @luciecharlesneuro.bsky.social and @wimdeneys.bsky.social, and to my lab @lapsyde.bsky.social.
Stay tuned for the computational modeling part! 🤓
You can access the preprint here: osf.io/preprints/ps...
OSF
osf.io
October 16, 2025 at 4:17 PM
10/10
Huge thanks to my great co-authors @zoepurcell.bsky.social , @luciecharlesneuro.bsky.social and @wimdeneys.bsky.social, and to my lab @lapsyde.bsky.social.
Stay tuned for the computational modeling part! 🤓
You can access the preprint here: osf.io/preprints/ps...
Huge thanks to my great co-authors @zoepurcell.bsky.social , @luciecharlesneuro.bsky.social and @wimdeneys.bsky.social, and to my lab @lapsyde.bsky.social.
Stay tuned for the computational modeling part! 🤓
You can access the preprint here: osf.io/preprints/ps...
They should do something about the acronym though ;-)
July 12, 2025 at 12:58 PM
They should do something about the acronym though ;-)
Goal: hopefully support more precise theorizing and empirical work on “System 2”—both in human and machine cognition. Full paper: osf.io/preprints/ps... (5/5)
OSF
osf.io
May 8, 2025 at 9:29 AM
Goal: hopefully support more precise theorizing and empirical work on “System 2”—both in human and machine cognition. Full paper: osf.io/preprints/ps... (5/5)
I propose guiding principles to avoid problematic misconceptions and point to critical outstanding issues we need to address in the coming years. (4/5)
May 8, 2025 at 9:29 AM
I propose guiding principles to avoid problematic misconceptions and point to critical outstanding issues we need to address in the coming years. (4/5)
Key point is that deliberation should be understood as multifunctional, serving multiple, complementary purposes. (3/5)
May 8, 2025 at 9:29 AM
Key point is that deliberation should be understood as multifunctional, serving multiple, complementary purposes. (3/5)
Although deliberation is central to dual-process theories, its conceptualization remains vague. I sketch a framework clarifying what System 2 deliberation is and does. (2/5)
May 8, 2025 at 9:29 AM
Although deliberation is central to dual-process theories, its conceptualization remains vague. I sketch a framework clarifying what System 2 deliberation is and does. (2/5)
It's been contested from the start :-) but good point: "examined" would be more accurate
March 4, 2025 at 9:26 AM
It's been contested from the start :-) but good point: "examined" would be more accurate
(6/6)
This preference can also be misused.
If people trust deliberation intuitively, it means both humans & AI can appear more trustworthy simply by framing decisions as “carefully reasoned”—even if they aren’t.
Read the full study: doi.org/10.31234/osf...
This preference can also be misused.
If people trust deliberation intuitively, it means both humans & AI can appear more trustworthy simply by framing decisions as “carefully reasoned”—even if they aren’t.
Read the full study: doi.org/10.31234/osf...
OSF
doi.org
February 18, 2025 at 4:24 PM
(6/6)
This preference can also be misused.
If people trust deliberation intuitively, it means both humans & AI can appear more trustworthy simply by framing decisions as “carefully reasoned”—even if they aren’t.
Read the full study: doi.org/10.31234/osf...
This preference can also be misused.
If people trust deliberation intuitively, it means both humans & AI can appear more trustworthy simply by framing decisions as “carefully reasoned”—even if they aren’t.
Read the full study: doi.org/10.31234/osf...
(5/6)
Why might this matter (even for "fast-and-slow" haters 😉)?
1️⃣ Deliberation shapes trust in advice
2️⃣ AI models already simulate deliberation (e.g., chain-of-thought reasoning).
3️⃣ Knowing this, AI developers can boost trust & fight algorithm aversion.
But…
Why might this matter (even for "fast-and-slow" haters 😉)?
1️⃣ Deliberation shapes trust in advice
2️⃣ AI models already simulate deliberation (e.g., chain-of-thought reasoning).
3️⃣ Knowing this, AI developers can boost trust & fight algorithm aversion.
But…
February 18, 2025 at 4:24 PM
(5/6)
Why might this matter (even for "fast-and-slow" haters 😉)?
1️⃣ Deliberation shapes trust in advice
2️⃣ AI models already simulate deliberation (e.g., chain-of-thought reasoning).
3️⃣ Knowing this, AI developers can boost trust & fight algorithm aversion.
But…
Why might this matter (even for "fast-and-slow" haters 😉)?
1️⃣ Deliberation shapes trust in advice
2️⃣ AI models already simulate deliberation (e.g., chain-of-thought reasoning).
3️⃣ Knowing this, AI developers can boost trust & fight algorithm aversion.
But…
(4/6)
Even under time pressure or cognitive load, participants still rated deliberation as better—ironically, indicating that the deliberation preference is itself intuitive.
Interestingly, ChatGPT (3.5 & 4) showed the same preference, suggesting AI models encode human folk beliefs about reasoning
Even under time pressure or cognitive load, participants still rated deliberation as better—ironically, indicating that the deliberation preference is itself intuitive.
Interestingly, ChatGPT (3.5 & 4) showed the same preference, suggesting AI models encode human folk beliefs about reasoning
February 18, 2025 at 4:24 PM
(4/6)
Even under time pressure or cognitive load, participants still rated deliberation as better—ironically, indicating that the deliberation preference is itself intuitive.
Interestingly, ChatGPT (3.5 & 4) showed the same preference, suggesting AI models encode human folk beliefs about reasoning
Even under time pressure or cognitive load, participants still rated deliberation as better—ironically, indicating that the deliberation preference is itself intuitive.
Interestingly, ChatGPT (3.5 & 4) showed the same preference, suggesting AI models encode human folk beliefs about reasoning
(3/6)
We asked participants to rate individuals who reasoned intuitively vs. deliberatively—while controlling whether the individual was portrayed as accurate or not.
Across all conditions, we find a strong preference for deliberation.
We asked participants to rate individuals who reasoned intuitively vs. deliberatively—while controlling whether the individual was portrayed as accurate or not.
Across all conditions, we find a strong preference for deliberation.
February 18, 2025 at 4:24 PM
(3/6)
We asked participants to rate individuals who reasoned intuitively vs. deliberatively—while controlling whether the individual was portrayed as accurate or not.
Across all conditions, we find a strong preference for deliberation.
We asked participants to rate individuals who reasoned intuitively vs. deliberatively—while controlling whether the individual was portrayed as accurate or not.
Across all conditions, we find a strong preference for deliberation.
(2/6)
While dual process research has pinpointed the mechanics of intuitive and deliberate thinking, we still lack understanding of how people perceive and value these modes of thought—what might be termed a “folk theory” of fast-and-slow thinking.
While dual process research has pinpointed the mechanics of intuitive and deliberate thinking, we still lack understanding of how people perceive and value these modes of thought—what might be termed a “folk theory” of fast-and-slow thinking.
February 18, 2025 at 4:24 PM
(2/6)
While dual process research has pinpointed the mechanics of intuitive and deliberate thinking, we still lack understanding of how people perceive and value these modes of thought—what might be termed a “folk theory” of fast-and-slow thinking.
While dual process research has pinpointed the mechanics of intuitive and deliberate thinking, we still lack understanding of how people perceive and value these modes of thought—what might be termed a “folk theory” of fast-and-slow thinking.