Maximilian Maier π
@MaxMa1er
Followers
549
Following
541
Media
30
Statuses
402
https://t.co/vu1jo3uzR8 PhD candidate @EP_UCL | Decision-making, applied statistics, & meta-analysis | Open science | Previously @ResMaPsychology
London
Joined November 2018
New Cogsci preprint, "Introducing the Extinction Gambling Task," with Adam Harris, David Kellen, and @HenrikSingmann. We introduce a risky choice task in which participants can lose their (entire) bonus pay when choosing the risky option and drawing the extinction event.
1
8
15
In this article, Maier et al. show that metacognitive learning from consequences can shape moral decision-making. @MaxMa1er @vanessachg_ @FalkLieder @RtnlAltruismLab
https://t.co/wnLzXdz7Mv
0
6
18
Excited to share that our paper is now published in PNAS!
pnas.org
As large language models (LLMs) become more widely used, people increasingly rely on them to make or advise on moral decisions. Some researchers ev...
We updated this preprint ( https://t.co/GriXMbkJBl) with a new study investigating sources of this yes-no framing bias and amplified omission bias in LLM moral decision-making. Results suggest that they likely arise from fine-tuning for chatbot applications. @MaxMa1er @FalkLieder
0
4
6
New preprint w/ @MaxMa1er, Harry Coulson & @FalkLieder: Telling people their global income rank (RBN) increased charitable donations. Also explored RBNβs effect on perceived global income distribution + whether asking how much others should donate increased giving (IDT) π§΅β¦
1
9
8
New preprint w/ @MaxMa1er, @vanessachg_, Fiery Cushman, & @FalkLieder: Disentangling Model-Based and Model-Free Moral Learning https://t.co/AnnnyYE0Se Excited to present this work at #CogSci2025!
1
6
8
There is a major improvement to the RoBMA package. Model structure and parameters can be estimated within a single MCMC yielding a tremendous speedup! A Bayesian publication bias model-averaged meta-regression can be estimated in less than a minute now! (down from 30min)
3
3
22
We updated this preprint ( https://t.co/GriXMbkJBl) with a new study investigating sources of this yes-no framing bias and amplified omission bias in LLM moral decision-making. Results suggest that they likely arise from fine-tuning for chatbot applications. @MaxMa1er @FalkLieder
New preprint with @MaxMa1er and @FalkLieder, titled "Large Language Models Amplify Human Biases in Moral Decision-Making." In this paper, LLMs and participants make decisions in realistic moral dilemmas. We found that LLMs exhibited stronger omission bias than participants.
0
6
9
For anyone who is still here, I am planning to mostly post on the other place from now on!
0
1
2
Our paper on "Model-Averaged Bayesian t tests" is now published in Psychonomic Bulletin and Review https://t.co/Th5Bm52YOS Tutorial video available at: https://t.co/ZfA0sejZJG Summary of the paper belowπ
π¨ New preprint "Model-Averaged Bayesian t-Tests" with @BartosFra, @dsquintana, @fdabl, @donvdbergh, @MarsmanMaarten, @AlexanderLyNL, & @EJWagenmakers! https://t.co/UIf8WVeImm
0
6
16
New Psychology Today post: "Failure to Replicate the Identified Victim Effect: We may need to rethink and reframe charitable-giving biases" Read: https://t.co/bJxiVuzIUI by Rajarshi Majumder, who led the project to a publication in JDM. Preprint: https://t.co/N6fBeAwSkl
1
17
37
π¨ Job alert π¨ One more week to apply for two positions (PhD and Postdoc) at @arc_mpib and @mpib_berlin on using LLMs to improve generalizability in the behavioral and social sciences LLMs. The positions are fully funded for three years. Apply by October 13! Thank you for
1
51
134
New paper! The Replication Database: Documenting the Replicability of Psychological Science. Check a novel interactive platform hosting 1,239 originals paired w/ replication findings https://t.co/XnhAvE4kZ1. Published in @up_jopd led by @aufdroeseler
https://t.co/S99gphdJ68
4
62
157
Upcoming hybrid talk at HKU: "Metacognitive Moral Learning and Decision Making in Realistic Moral Dilemmas" by Max Maier & Vanessa Cheung from UCL Sept 13 1230β200pm HK time (GMT+8) HKU, Jockey Club Tower, 813 or join us on Zoom: https://t.co/llTVBoYIqp (pass: psyc)
1
6
7
π¨New preprint!π¨Excited to share my 1st PhD project with @BonanZhao @cocosci_lab @natvelali Teaching is a powerful way to pass on "tools for thought" to solve new problems. How well do existing teaching models of teaching capture this process? Link:
3
18
68
Using this design, we again replicated the transfer effects on moral convictions and donation decisions between the two independent studies. For more details see our preprint: https://t.co/jOkh6ArxUr
0
0
0
Lastly, to address potential demand characteristics, we recruited participants for a study with the learning task and then recruited the same participants for a second study with the transfer measures. Ps did not know that the two studies were related (more details in Methods).
1
0
0
The learning of those participants that are best captured by metacognitive learning transferred to incentive-compatible donation decisions and measures of moral convictions.
1
0
0
Further, using an ensemble of computational models, we show that many participants learn about decision strategies in general (metacognitive learning) rather than specific behaviors.
1
0
0
Across four experiments, we found that participants adjusted their decision-making strategies (reliance on rules vs. cost-benefit reasoning) based on which produced better consequences.
1
0
0
We outline a theory of how strategy selection learning shapes moral decision-making, implement it using computational models, and test it using a novel paradigm involving realistic trolley-type dilemmas with outcomes.
1
0
0