diff --git a/content/publication/franken2023rails.md b/content/publication/franken2023rails.md
new file mode 100644
index 0000000..a2f8185
--- /dev/null
+++ b/content/publication/franken2023rails.md
@@ -0,0 +1,33 @@
++++
+# 0 -> 'Forthcoming',
+# 1 -> 'Preprint',
+# 2 -> 'Journal',
+# 3 -> 'Conference Proceedings',
+# 4 -> 'Book chapter',
+# 5 -> 'Thesis'
+
+title = "Off The Rails: Procedural Dilemma Generation for Moral Reasoning"
+date = "2023-10-30"
+authors = ['J. Fränken',"A. Khawaja","K. Gandhi","J. Moore","N. D. Goodman","T. Gerstenberg"]
+publication_types = ["3"]
+publication_short = "_AI Meets Moral Philosophy and Moral Psychology Workshop (NeurIPS 2023)_"
+publication = "Fränken J., Khawaja A., Gandhi K., Moore J., Goodman N. D., Gerstenberg T. (2023). Off The Rails: Procedural Dilemma Generation for Moral Reasoning. In _AI Meets Moral Philosophy and Moral Psychology Workshop (NeurIPS 2023)_."
+abstract = "As AI systems like language models are increasingly integrated into making decisions that affect people, it's critical to ensure that these systems have sound moral reasoning. To test whether they do, we need to develop systematic evaluations. Recent work has introduced a method for procedurally generating LLM evaluations from abstract causal templates, and tested this method in the context of social reasoning (i.e., theory-of-mind). In this paper, we extend this method to the domain of moral dilemmas. We develop a framework that translates causal graphs into a prompt template which can then be used to procedurally generate a large and diverse set of moral dilemmas using a language model. Using this framework, we created the OffTheRails dataset which consists of 50 scenarios and 500 unique test items. We evaluated the quality of our model-written test items using two independent human experts and found that 90% of the test-items met the desired structure. We collect moral permissibility and intention judgments from 100 human crowdworkers and compared these judgments with those from GPT-4 and Claude-2 across eight control conditions. Both humans and GPT-4 assigned higher intentionality to agents when a harmful outcome was evitable and a necessary means. However, our findings did not match previous findings on permissibility judgments. This difference may be a result of not controlling the severity of harmful outcomes during scenario generation. We conclude by discussing future extensions of our benchmark to address this limitation."
+image_preview = ""
+selected = false
+projects = []
+#url_pdf = "papers/franken2023rails.pdf"
+url_preprint = ""
+url_code = ""
+url_dataset = ""
+url_slides = ""
+url_video = ""
+url_poster = ""
+url_source = ""
+#url_custom = [{name = "Github", url = ""}]
+math = true
+highlight = true
+[header]
+# image = "publications/franken2023rails.png"
+caption = ""
++++
\ No newline at end of file
diff --git a/content/publication/franken2023social.md b/content/publication/franken2023social.md
new file mode 100644
index 0000000..f9834d5
--- /dev/null
+++ b/content/publication/franken2023social.md
@@ -0,0 +1,33 @@
++++
+# 0 -> 'Forthcoming',
+# 1 -> 'Preprint',
+# 2 -> 'Journal',
+# 3 -> 'Conference Proceedings',
+# 4 -> 'Book chapter',
+# 5 -> 'Thesis'
+
+title = "Social Contract AI: Aligning AI Assistants with Implicit Group Norms"
+date = "2023-10-30"
+authors = ['J. Fränken',"S. Kwok","P. Ye","K. Gandhi","D. Arumugam","J. Moore","A. Tamkin","T. Gerstenberg","N. D. Goodman"]
+publication_types = ["3"]
+publication_short = "_Socially Responsible Language Modelling Research Workshop (NeurIPS 2023)_"
+publication = "Fränken J., Kwok S., Ye P., Gandhi K., Arumugam D., Moore J., Tamkin A., Gerstenberg T., Goodman N. D. (2023). Social Contract AI: Aligning AI Assistants with Implicit Group Norms. In _Socially Responsible Language Modelling Research Workshop (NeurIPS 2023)."
+abstract = "We explore the idea of aligning an AI assistant by inverting a model of users' (unknown) preferences from observed interactions. To validate our proposal, we run proof-of-concept simulations in the economic ultimatum game, formalizing user preferences as policies that guide the actions of simulated players. We find that the AI assistant accurately aligns its behavior to match standard policies from the economic literature (e.g., selfish, altruistic). However, the assistant's learned policies lack robustness and exhibit limited generalization in an out-of-distribution setting when confronted with a currency (e.g., grams of medicine) that was not included in the assistant's training distribution. Additionally, we find that when there is inconsistency in the relationship between language use and an unknown policy (e.g., an altruistic policy combined with rude language), the assistant's learning of the policy is slowed. Overall, our preliminary results suggest that developing simulation frameworks in which AI assistants need to infer preferences from diverse users can provide a valuable approach for studying practical alignment questions."
+image_preview = ""
+selected = false
+projects = []
+url_pdf = "papers/franken2023social.pdf"
+url_preprint = "https://arxiv.org/abs/2310.17769"
+url_code = ""
+url_dataset = ""
+url_slides = ""
+url_video = ""
+url_poster = ""
+url_source = ""
+url_custom = [{name = "Github", url = "https://github.com/janphilippfranken/scai/tree/release"}]
+math = true
+highlight = true
+[header]
+# image = "publications/franken2023social.png"
+caption = ""
++++
\ No newline at end of file
diff --git a/content/publication/gandhi2023understanding.md b/content/publication/gandhi2023understanding.md
index eaa0eac..f26adb0 100644
--- a/content/publication/gandhi2023understanding.md
+++ b/content/publication/gandhi2023understanding.md
@@ -9,9 +9,9 @@
title = "Understanding Social Reasoning in Language Models with Language Models"
date = "2023-06-27"
authors = ["K. Gandhi",'J. Fränken',"T. Gerstenberg","N. D. Goodman"]
-publication_types = ["1"]
-publication_short = "_arXiv_"
-publication = "Gandhi K., Fränken J., Gerstenberg T., Goodman N. D. (2023). Understanding Social Reasoning in Language Models with Language Models. In _arXiv_."
+publication_types = ["3"]
+publication_short = "_Advances in Neural Information Processing Systems_"
+publication = "Gandhi K., Fränken J., Gerstenberg T., Goodman N. D. (2023). Understanding Social Reasoning in Language Models with Language Models. _Advances in Neural Information Processing Systems_."
abstract = "As Large Language Models (LLMs) become increasingly integrated into our everyday lives, understanding their ability to comprehend human mental states becomes critical for ensuring effective interactions. However, despite the recent attempts to assess the Theory-of-Mind (ToM) reasoning capabilities of LLMs, the degree to which these models can align with human ToM remains a nuanced topic of exploration. This is primarily due to two distinct challenges: (1) the presence of inconsistent results from previous evaluations, and (2) concerns surrounding the validity of existing evaluation methodologies. To address these challenges, we present a novel framework for procedurally generating evaluations with LLMs by populating causal templates. Using our framework, we create a new social reasoning benchmark (BigToM) for LLMs which consists of 25 controls and 5,000 model-written evaluations. We find that human participants rate the quality of our benchmark higher than previous crowd-sourced evaluations and comparable to expert-written evaluations. Using BigToM, we evaluate the social reasoning capabilities of a variety of LLMs and compare model performances with human performance. Our results suggest that GPT4 has ToM capabilities that mirror human inference patterns, though less reliable, while other LLMs struggle."
tags = ["Spotlight at NeurIPS 2023"]
image_preview = ""
diff --git a/content/publication/goodman2023probabilistic.md b/content/publication/goodman2023probabilistic.md
index 11ad464..b41caae 100644
--- a/content/publication/goodman2023probabilistic.md
+++ b/content/publication/goodman2023probabilistic.md
@@ -7,7 +7,7 @@
# 5 -> 'Thesis'
title = "Probabilistic programs as a unifying language of thought"
-date = "2023-01-01"
+date = "2023-10-20"
year = "{in press}"
authors = ["N. D. Goodman","T. Gerstenberg","J. B. Tenenbaum"]
publication_types = ["4", "0"]
diff --git a/content/publication/kirfel2023anticipating.md b/content/publication/kirfel2023anticipating.md
new file mode 100644
index 0000000..1989749
--- /dev/null
+++ b/content/publication/kirfel2023anticipating.md
@@ -0,0 +1,33 @@
++++
+# 0 -> 'Forthcoming',
+# 1 -> 'Preprint',
+# 2 -> 'Journal',
+# 3 -> 'Conference Proceedings',
+# 4 -> 'Book chapter',
+# 5 -> 'Thesis'
+
+title = "Anticipating the risks and benefits of counterfactual world simulation models"
+date = "2023-10-30"
+authors = ["L. Kirfel","R. J. MacCoun","T. Icard","T. Gerstenberg"]
+publication_types = ["3"]
+publication_short = "_AI Meets Moral Philosophy and Moral Psychology Workshop (NeurIPS 2023)_"
+publication = "Kirfel L., MacCoun R. J., Icard T., Gerstenberg T. (2023). Anticipating the risks and benefits of counterfactual world simulation models. In _AI Meets Moral Philosophy and Moral Psychology Workshop (NeurIPS 2023)_."
+abstract = "This paper examines the transformative potential of Counterfactual World Simulation Models (CWSMs). A CWSM uses multi-modal evidence, such as the CCTV footage of a road accident, to build a high-fidelity 3D reconstruction of what happened. It can answer causal questions, such as whether the accident happened because the driver was speeding, by simulating what would have happened in relevant counterfactual situations. We argue for a normative and ethical framework that guides and constrains the simulation of counterfactuals. We address the challenge of ensuring fidelity in reconstructions while simultaneously preventing stereotype perpetuation during counterfactual simulations. We anticipate different modes of how users will interact with CWSMs and discuss how their outputs may be presented. Finally, we address the prospective applications of CWSMs in the legal domain, recognizing both their potential to revolutionize legal proceedings as well as the ethical concerns they engender. Sketching a new genre of AI, this paper seeks to illuminate the path forward for responsible and effective use of CWSMs."
+image_preview = ""
+selected = false
+projects = []
+#url_pdf = "papers/kirfel2023anticipating.pdf"
+url_preprint = ""
+url_code = ""
+url_dataset = ""
+url_slides = ""
+url_video = ""
+url_poster = ""
+url_source = ""
+#url_custom = [{name = "Github", url = ""}]
+math = true
+highlight = true
+[header]
+# image = "publications/kirfel2023anticipating.png"
+caption = ""
++++
\ No newline at end of file
diff --git a/docs/404.html b/docs/404.html
index 92cb07e..663093b 100644
--- a/docs/404.html
+++ b/docs/404.html
@@ -238,23 +238,23 @@
+
+ J. Fränken, S. Kwok, P. Ye, K. Gandhi, D. Arumugam, J. Moore, A. Tamkin, T. Gerstenberg, N. D. Goodman
+
+ (2023).
+
+ Social Contract AI: Aligning AI Assistants with Implicit Group Norms.
+ Socially Responsible Language Modelling Research Workshop (NeurIPS 2023).
+
+
+
+
+
Off The Rails: Procedural Dilemma Generation for Moral Reasoning
+
+
+ J. Fränken, A. Khawaja, K. Gandhi, J. Moore, N. D. Goodman, T. Gerstenberg
+
+
+
+
+
+
+
+
+
+
+
+
Abstract
+
As AI systems like language models are increasingly integrated into making decisions that affect people, it’s critical to ensure that these systems have sound moral reasoning. To test whether they do, we need to develop systematic evaluations. Recent work has introduced a method for procedurally generating LLM evaluations from abstract causal templates, and tested this method in the context of social reasoning (i.e., theory-of-mind). In this paper, we extend this method to the domain of moral dilemmas. We develop a framework that translates causal graphs into a prompt template which can then be used to procedurally generate a large and diverse set of moral dilemmas using a language model. Using this framework, we created the OffTheRails dataset which consists of 50 scenarios and 500 unique test items. We evaluated the quality of our model-written test items using two independent human experts and found that 90% of the test-items met the desired structure. We collect moral permissibility and intention judgments from 100 human crowdworkers and compared these judgments with those from GPT-4 and Claude-2 across eight control conditions. Both humans and GPT-4 assigned higher intentionality to agents when a harmful outcome was evitable and a necessary means. However, our findings did not match previous findings on permissibility judgments. This difference may be a result of not controlling the severity of harmful outcomes during scenario generation. We conclude by discussing future extensions of our benchmark to address this limitation.
Fränken J., Khawaja A., Gandhi K., Moore J., Goodman N. D., Gerstenberg T. (2023). Off The Rails: Procedural Dilemma Generation for Moral Reasoning. In AI Meets Moral Philosophy and Moral Psychology Workshop (NeurIPS 2023).
Social Contract AI: Aligning AI Assistants with Implicit Group Norms
+
+
+ J. Fränken, S. Kwok, P. Ye, K. Gandhi, D. Arumugam, J. Moore, A. Tamkin, T. Gerstenberg, N. D. Goodman
+
+
+
+
+
+
+
+
+
+
+
+
Abstract
+
We explore the idea of aligning an AI assistant by inverting a model of users’ (unknown) preferences from observed interactions. To validate our proposal, we run proof-of-concept simulations in the economic ultimatum game, formalizing user preferences as policies that guide the actions of simulated players. We find that the AI assistant accurately aligns its behavior to match standard policies from the economic literature (e.g., selfish, altruistic). However, the assistant’s learned policies lack robustness and exhibit limited generalization in an out-of-distribution setting when confronted with a currency (e.g., grams of medicine) that was not included in the assistant’s training distribution. Additionally, we find that when there is inconsistency in the relationship between language use and an unknown policy (e.g., an altruistic policy combined with rude language), the assistant’s learning of the policy is slowed. Overall, our preliminary results suggest that developing simulation frameworks in which AI assistants need to infer preferences from diverse users can provide a valuable approach for studying practical alignment questions.
Fränken J., Kwok S., Ye P., Gandhi K., Arumugam D., Moore J., Tamkin A., Gerstenberg T., Goodman N. D. (2023). Social Contract AI: Aligning AI Assistants with Implicit Group Norms. In _Socially Responsible Language Modelling Research Workshop (NeurIPS 2023).
Gandhi K., Fränken J., Gerstenberg T., Goodman N. D. (2023). Understanding Social Reasoning in Language Models with Language Models. In arXiv.
+
Gandhi K., Fränken J., Gerstenberg T., Goodman N. D. (2023). Understanding Social Reasoning in Language Models with Language Models. Advances in Neural Information Processing Systems.
+
+ J. Fränken, S. Kwok, P. Ye, K. Gandhi, D. Arumugam, J. Moore, A. Tamkin, T. Gerstenberg, N. D. Goodman
+
+ (2023).
+
+ Social Contract AI: Aligning AI Assistants with Implicit Group Norms.
+ Socially Responsible Language Modelling Research Workshop (NeurIPS 2023).
+
+
+
+
+
diff --git a/docs/publication/index.xml b/docs/publication/index.xml
index 7165e47..16503fa 100644
--- a/docs/publication/index.xml
+++ b/docs/publication/index.xml
@@ -12,6 +12,33 @@
+
+ Anticipating the risks and benefits of counterfactual world simulation models
+ https://cicl.stanford.edu/publication/kirfel2023anticipating/
+ Mon, 30 Oct 2023 00:00:00 +0000
+
+ https://cicl.stanford.edu/publication/kirfel2023anticipating/
+
+
+
+
+ Off The Rails: Procedural Dilemma Generation for Moral Reasoning
+ https://cicl.stanford.edu/publication/franken2023rails/
+ Mon, 30 Oct 2023 00:00:00 +0000
+
+ https://cicl.stanford.edu/publication/franken2023rails/
+
+
+
+
+ Social Contract AI: Aligning AI Assistants with Implicit Group Norms
+ https://cicl.stanford.edu/publication/franken2023social/
+ Mon, 30 Oct 2023 00:00:00 +0000
+
+ https://cicl.stanford.edu/publication/franken2023social/
+
+
+
If not me, then who? Responsibility and replacement
https://cicl.stanford.edu/publication/wu2023replacement/
@@ -21,6 +48,15 @@
+
+ Probabilistic programs as a unifying language of thought
+ https://cicl.stanford.edu/publication/goodman2023probabilistic/
+ Fri, 20 Oct 2023 00:00:00 +0000
+
+ https://cicl.stanford.edu/publication/goodman2023probabilistic/
+
+
+
Children use disagreement to infer what happened
https://cicl.stanford.edu/publication/amemiya2023disagreement/
@@ -174,15 +210,6 @@
-
- Probabilistic programs as a unifying language of thought
- https://cicl.stanford.edu/publication/goodman2023probabilistic/
- Sun, 01 Jan 2023 00:00:00 +0000
-
- https://cicl.stanford.edu/publication/goodman2023probabilistic/
-
-
-
What would have happened? Counterfactuals, hypotheticals, and causal judgments
https://cicl.stanford.edu/publication/gerstenberg2022hypothetical/
diff --git a/docs/publication/kirfel2023anticipating/index.html b/docs/publication/kirfel2023anticipating/index.html
new file mode 100644
index 0000000..8f555f6
--- /dev/null
+++ b/docs/publication/kirfel2023anticipating/index.html
@@ -0,0 +1,484 @@
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ Anticipating the risks and benefits of counterfactual world simulation models | Causality in Cognition Lab
+
+
+
+
+
+
+
+
+
+
+
+
+
+
Anticipating the risks and benefits of counterfactual world simulation models
+
+
+ L. Kirfel, R. J. MacCoun, T. Icard, T. Gerstenberg
+
+
+
+
+
+
+
+
+
+
+
+
Abstract
+
This paper examines the transformative potential of Counterfactual World Simulation Models (CWSMs). A CWSM uses multi-modal evidence, such as the CCTV footage of a road accident, to build a high-fidelity 3D reconstruction of what happened. It can answer causal questions, such as whether the accident happened because the driver was speeding, by simulating what would have happened in relevant counterfactual situations. We argue for a normative and ethical framework that guides and constrains the simulation of counterfactuals. We address the challenge of ensuring fidelity in reconstructions while simultaneously preventing stereotype perpetuation during counterfactual simulations. We anticipate different modes of how users will interact with CWSMs and discuss how their outputs may be presented. Finally, we address the prospective applications of CWSMs in the legal domain, recognizing both their potential to revolutionize legal proceedings as well as the ethical concerns they engender. Sketching a new genre of AI, this paper seeks to illuminate the path forward for responsible and effective use of CWSMs.
Kirfel L., MacCoun R. J., Icard T., Gerstenberg T. (2023). Anticipating the risks and benefits of counterfactual world simulation models. In AI Meets Moral Philosophy and Moral Psychology Workshop (NeurIPS 2023).
-
- As Large Language Models (LLMs) become increasingly integrated into our everyday lives, understanding their ability to comprehend human mental states becomes critical for ensuring effective interactions. However, despite the recent attempts to assess …
-
-
- Interacting in real environments, such as manipulating objects, involves multisensory information. However, little is known about how multisensory cue characteristics help us determine what has occurred in a scene, including whether two events were …
+ This paper examines the transformative potential of Counterfactual World Simulation Models (CWSMs). A CWSM uses multi-modal evidence, such as the CCTV footage of a road accident, to build a high-fidelity 3D reconstruction of what happened. It can …
- How responsible someone is for an outcome depends on both the causal role of their actions, and what those actions reveal about their moral character. Prior work has successfully modeled people's causal attributions and mental state inferences using …
+ As AI systems like language models are increasingly integrated into making decisions that affect people, it's critical to ensure that these systems have sound moral reasoning. To test whether they do, we need to develop systematic evaluations. Recent …
- What someone knew matters for how we hold them responsible. In three studies, we explore people's responsibility judgments for negative outcomes to knowledgeable versus ignorant agents. We manipulate whether agents arrived at their knowledge state …
+ We explore the idea of aligning an AI assistant by inverting a model of users' (unknown) preferences from observed interactions. To validate our proposal, we run proof-of-concept simulations in the economic ultimatum game, formalizing user …
- There are at least three ways of learning how the world works: learning from observations, from interventions, and from explanations. Prior work on causal inference focused on how people learn causal structures through observation and intervention. …
+ Interacting in real environments, such as manipulating objects, involves multisensory information. However, little is known about how multisensory cue characteristics help us determine what has occurred in a scene, including whether two events were …
- How do essentialist beliefs about categories arise? We hypothesize that such beliefs are transmitted via language. We subject large language models (LLMs) to vignettes from the literature on essentialist categorization and find that they align well …
+ As Large Language Models (LLMs) become increasingly integrated into our everyday lives, understanding their ability to comprehend human mental states becomes critical for ensuring effective interactions. However, despite the recent attempts to assess …
- When choosing how to describe what happened, we have a number of causal verbs at our disposal. In this paper, we develop a model-theoretic formal semantics for nine causal verbs that span the categories of CAUSE, ENABLE, and PREVENT. We use …
+ How responsible someone is for an outcome depends on both the causal role of their actions, and what those actions reveal about their moral character. Prior work has successfully modeled people's causal attributions and mental state inferences using …
- This work attempts to bridge the divide between accounts of causal reasoning with respect to agents and objects. We begin by examining the influence of animacy. In a collision-based context, we vary the animacy status of an object using 3D …
+ What someone knew matters for how we hold them responsible. In three studies, we explore people's responsibility judgments for negative outcomes to knowledgeable versus ignorant agents. We manipulate whether agents arrived at their knowledge state …
- What shape do people's mental models take? We hypothesize that people build causal models that are suited to the task at hand. These models abstract away information to represent what matters. To test this idea empirically, we presented participants …
+ There are at least three ways of learning how the world works: learning from observations, from interventions, and from explanations. Prior work on causal inference focused on how people learn causal structures through observation and intervention. …
- Generic statements, such as "Bees are striped" are thought to be a central vehicle by which essentialist beliefs are transmitted. But work on generics and essentialism almost never focuses on the type of properties mentioned in generic statements. We …
+ How do essentialist beliefs about categories arise? We hypothesize that such beliefs are transmitted via language. We subject large language models (LLMs) to vignettes from the literature on essentialist categorization and find that they align well …
- Prior work has identified a resilient phenomenon that threatens the performance of human-AI decision-making teams: overreliance, when people agree with an AI, even when it is incorrect. Surprisingly, overreliance does not reduce when the AI produces …
+ When choosing how to describe what happened, we have a number of causal verbs at our disposal. In this paper, we develop a model-theoretic formal semantics for nine causal verbs that span the categories of CAUSE, ENABLE, and PREVENT. We use …
- Human adults can figure out what happened by combining evidence from different sensory modalities, such as vision and sound. How does the ability to integrate multi-modal information develop in early childhood? Inspired by prior computational work …
+ This work attempts to bridge the divide between accounts of causal reasoning with respect to agents and objects. We begin by examining the influence of animacy. In a collision-based context, we vary the animacy status of an object using 3D …
- How do people make causal judgments about other's decisions? Prior work has argued that judging causation requires going beyond what actually happened and simulating what would have happened in a relevant counterfactual situation. Here, we extend the …
+ What shape do people's mental models take? We hypothesize that people build causal models that are suited to the task at hand. These models abstract away information to represent what matters. To test this idea empirically, we presented participants …
- Mental simulation is a powerful cognitive capacity that underlies people's ability to draw inferences about what happened in the past from the present. Recent work suggests that eye-tracking can be used as a window through which one can study the …
+ Generic statements, such as "Bees are striped" are thought to be a central vehicle by which essentialist beliefs are transmitted. But work on generics and essentialism almost never focuses on the type of properties mentioned in generic statements. We …
- In many practical applications of AI, an AI model is used as a decision aid for human users. The AI provides advice that a human (sometimes) incorporates into their decision-making process. The AI advice is often presented with some measure of …
+ Prior work has identified a resilient phenomenon that threatens the performance of human-AI decision-making teams: overreliance, when people agree with an AI, even when it is incorrect. Surprisingly, overreliance does not reduce when the AI produces …
- Humans have a remarkable ability to go beyond the observable. From seeing the current state of our shared kitchen, we can infer what happened and who did it. Prior work has shown how the physical state of the world licenses inferences about the …
+ Human adults can figure out what happened by combining evidence from different sensory modalities, such as vision and sound. How does the ability to integrate multi-modal information develop in early childhood? Inspired by prior computational work …
- People use varied language to express their causal understanding of the world. But how does that language map onto people’s underlying representations, and how do people choose between competing ways to best describe what happened? In this paper we …
+ How do people make causal judgments about other's decisions? Prior work has argued that judging causation requires going beyond what actually happened and simulating what would have happened in a relevant counterfactual situation. Here, we extend the …
- To evaluate others' actions, we consider action outcomes (e.g., positive or negative) and the actors' underlying intentions (e.g., intentional or accidental). However, we often encounter situ- ations where neither actual outcomes nor intentions …
+ Mental simulation is a powerful cognitive capacity that underlies people's ability to draw inferences about what happened in the past from the present. Recent work suggests that eye-tracking can be used as a window through which one can study the …
- How do we estimate the difficulty of performing a new task, a task we've never tried before such as making a sculpture, a birthday cake, or building a tower with LEGO blocks? Estimating difficulty helps us appreciate others' accomplishments, and …
+ In many practical applications of AI, an AI model is used as a decision aid for human users. The AI provides advice that a human (sometimes) incorporates into their decision-making process. The AI advice is often presented with some measure of …
- Language that describes people in a concise manner may conflict with social norms (e.g., referring to people by their race), presenting a conflict between transferring information efficiently and avoiding offensive language. When a speaker is …
+ Humans have a remarkable ability to go beyond the observable. From seeing the current state of our shared kitchen, we can infer what happened and who did it. Prior work has shown how the physical state of the world licenses inferences about the …
- We introduce a novel experimental paradigm for studying multi-modal integration in causal inference. Our experiments feature a physically realistic Plinko machine in which a ball is dropped through one of three holes and comes to rest at the bottom …
+ People use varied language to express their causal understanding of the world. But how does that language map onto people’s underlying representations, and how do people choose between competing ways to best describe what happened? In this paper we …
- Event timing and interventions are important and intertwined cues to causal structure, yet they have typically been studied separately. We bring them together for the first time in an experiment where participants learn causal structure by performing …
+ To evaluate others' actions, we consider action outcomes (e.g., positive or negative) and the actors' underlying intentions (e.g., intentional or accidental). However, we often encounter situ- ations where neither actual outcomes nor intentions …
- In this paper we introduce the hypothetical simulation model (HSM) of physical support. The HSM predicts that people judge physical support by mentally simulating what would happen if the object of interest were removed. Two experiments test the …
+ How do we estimate the difficulty of performing a new task, a task we've never tried before such as making a sculpture, a birthday cake, or building a tower with LEGO blocks? Estimating difficulty helps us appreciate others' accomplishments, and …
- Consider the following causal explanation: The ball went through the goal because the defender didn’t block it. There are at least two problems with citing omissions as causal explanations. First, how do we choose the relevant candidate omission …
+ Language that describes people in a concise manner may conflict with social norms (e.g., referring to people by their race), presenting a conflict between transferring information efficiently and avoiding offensive language. When a speaker is …
- In this paper, we present a new task that investigates how people interact with and make judgments about towers of blocks. In Experiment 1, participants in the lab solved a series of problems in which they had to re-configure three blocks from an …
+ We introduce a novel experimental paradigm for studying multi-modal integration in causal inference. Our experiments feature a physically realistic Plinko machine in which a ball is dropped through one of three holes and comes to rest at the bottom …
- Moral judgment often involves pinning causation for harm to a particular person. Since it reveals “who one sides with”, expression of moral judgment can be a costly social act that people may be motivated to conceal. Here, we demonstrate that a …
+ Event timing and interventions are important and intertwined cues to causal structure, yet they have typically been studied separately. We bring them together for the first time in an experiment where participants learn causal structure by performing …
- In this paper, we bring together research on active learning and intuitive physics to explore how people learn about “microworlds” with continuous spatiotemporal dynamics. Participants interacted with objects in simple two-dimensional worlds governed …
+ In this paper we introduce the hypothetical simulation model (HSM) of physical support. The HSM predicts that people judge physical support by mentally simulating what would happen if the object of interest were removed. Two experiments test the …
- When did something almost happen? In this paper, we investigate what brings counterfactual worlds close. In Experiments 1 and 2, we find that participants’ judgments about whether something almost happened are determined by the causal proximity of …
+ Consider the following causal explanation: The ball went through the goal because the defender didn’t block it. There are at least two problems with citing omissions as causal explanations. First, how do we choose the relevant candidate omission …
- Many social judgments hinge on assigning responsibility to individuals for their role in a group’s success or failure. Often the group’s success depends on every team member acting in a rational way. When someone does not conform to what others …
+ In this paper, we present a new task that investigates how people interact with and make judgments about towers of blocks. In Experiment 1, participants in the lab solved a series of problems in which they had to re-configure three blocks from an …
- How do people make causal judgments? Here, we propose a counterfactual simulation model (CSM) of causal judgment that unifies different views on causation. The CSM predicts that people’s causal judgments are influenced by whether a candidate cause …
+ Moral judgment often involves pinning causation for harm to a particular person. Since it reveals “who one sides with”, expression of moral judgment can be a costly social act that people may be motivated to conceal. Here, we demonstrate that a …
- The actions of a rational agent reveal information about its mental states. These inferred mental states, particularly the agent’s intentions, play an important role in the evaluation of moral permissibility. While previous computational models have …
+ In this paper, we bring together research on active learning and intuitive physics to explore how people learn about “microworlds” with continuous spatiotemporal dynamics. Participants interacted with objects in simple two-dimensional worlds governed …
- How do people assign responsibility for the outcome of an election? In previous work, we have shown that responsibility judgments in achievement contexts are affected by the probability that a person’s contribution is necessary, and by how close it …
+ When did something almost happen? In this paper, we investigate what brings counterfactual worlds close. In Experiments 1 and 2, we find that participants’ judgments about whether something almost happened are determined by the causal proximity of …
- When agents violate norms, they are typically judged to be more of a cause of resulting outcomes. In this study, we suggest that norm violations also reduce the causality of other agents, a novel phenomenon we refer to as “causal supersession.” We …
+ Many social judgments hinge on assigning responsibility to individuals for their role in a group’s success or failure. Often the group’s success depends on every team member acting in a rational way. When someone does not conform to what others …
- In this paper, we demonstrate that people’s causal judgments are inextricably linked to counterfactuals. In our experiments, participants judge whether one billiard ball A caused another ball B to go through a gate. Our counterfactual simulation …
+ How do people make causal judgments? Here, we propose a counterfactual simulation model (CSM) of causal judgment that unifies different views on causation. The CSM predicts that people’s causal judgments are influenced by whether a candidate cause …
- The timing and order in which a set of events occur strongly influences whether people judge them to be causally related. But what do people think particular temporal patterns of events tell them about causal structure? And how do they integrate …
+ The actions of a rational agent reveal information about its mental states. These inferred mental states, particularly the agent’s intentions, play an important role in the evaluation of moral permissibility. While previous computational models have …
- In order to be held responsible, a person’s action has to have made some sort of difference to the outcome. In this paper, we propose a counterfactual replacement model according to which people attribute responsibility by comparing their prior …
+ How do people assign responsibility for the outcome of an election? In previous work, we have shown that responsibility judgments in achievement contexts are affected by the probability that a person’s contribution is necessary, and by how close it …
- Would Dan have died if Bob hadn’t shot? In this paper, we show that people’s answer depends on whether or not they are asked about what would have caused Bob not to shoot. Something needs to change in order to turn an actual world into a …
+ When agents violate norms, they are typically judged to be more of a cause of resulting outcomes. In this study, we suggest that norm violations also reduce the causality of other agents, a novel phenomenon we refer to as “causal supersession.” We …
- There is a long tradition in both philosophy and psychology to separate process accounts from dependency accounts of causation. In this paper, we motivate a unifying account that explains people’s causal attributions in terms of counterfactuals …
+ In this paper, we demonstrate that people’s causal judgments are inextricably linked to counterfactuals. In our experiments, participants judge whether one billiard ball A caused another ball B to go through a gate. Our counterfactual simulation …
- How do people make inferences from complex patterns of evidence across diverse situations? What does a computational model need in order to capture the abstract knowledge people use for everyday reasoning? In this paper, we explore a novel modeling …
+ The timing and order in which a set of events occur strongly influences whether people judge them to be causally related. But what do people think particular temporal patterns of events tell them about causal structure? And how do they integrate …
- We consider an approach to blame attribution based on counterfactual reasoning in probabilistic generative models. In this view, people intervene on each variable within their model and assign blame in proportion to how much a change to a variable …
+ In order to be held responsible, a person’s action has to have made some sort of difference to the outcome. In this paper, we propose a counterfactual replacement model according to which people attribute responsibility by comparing their prior …
- To what extent do people care about the intentions behind an action? What if the intentions can be deceptive? We conducted two experiments to complement previous evidence about the roles of outcomes and intentions in economic games. The results of …
+ Would Dan have died if Bob hadn’t shot? In this paper, we show that people’s answer depends on whether or not they are asked about what would have caused Bob not to shoot. Something needs to change in order to turn an actual world into a …
+
+ There is a long tradition in both philosophy and psychology to separate process accounts from dependency accounts of causation. In this paper, we motivate a unifying account that explains people’s causal attributions in terms of counterfactuals …
+
+
+
+ How do people make inferences from complex patterns of evidence across diverse situations? What does a computational model need in order to capture the abstract knowledge people use for everyday reasoning? In this paper, we explore a novel modeling …
+
+
+
+ We consider an approach to blame attribution based on counterfactual reasoning in probabilistic generative models. In this view, people intervene on each variable within their model and assign blame in proportion to how much a change to a variable …
+
+
+
+ To what extent do people care about the intentions behind an action? What if the intentions can be deceptive? We conducted two experiments to complement previous evidence about the roles of outcomes and intentions in economic games. The results of …
+
+