How Humans Perceive Human-like Behavior in Video Game Navigation

Zuniga, Evelyn; Milani, Stephanie; Leroy, Guy; Rzepecki, Jaroslaw; Georgescu, Raluca; Momennejad, Ida; Bignell, Dave; Sun, Mingfei; Shaw, Alison; Costello, Gavin; Jacob, Mikhail; Devlin, Sam; Hofmann, Katja

doi:10.1145/3491101.3519735

Cited by 4 publications

(9 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…rat navigation, [6]; on the use of Turing Tests, e.g. the Navigation Turing Test (NTT) for 3D video games [1][2][3]; or on deep learning and other machine learning approaches to analysing or generating behaviour (e.g. for intuitive physics [109]).…”

Section: What We Talk About When We Talk About Human-like Behaviourmentioning

confidence: 99%

“…The study [1] and its replication [2] revealed that the two SoTA agents [116], with the same performance on previous navigation benchmarks, were not judged equally human-like by human judges, and neither passed the NTT (figure 3). In addition, six artificial agents with different architectures and training measures were designed to judge the human-likeness of videos as well, a test called Artificial NTT (ANTT).…”

Section: (E) Human Navigation Turing Testmentioning

confidence: 99%

“…While this framework was mostly inspired by early twentieth-century psychology of trial and error behaviour, as well as later theories of associative learning and neural computation [88,89], recent advances in RL capture learning cognitive maps, generalization and transfer, metalearning, and learning structures and representations [90]. Some RL algorithms are tested as hypotheses of neural representations underpinning underlying human behaviour [54,90], behaviourally successful RL algorithms are tested as hypotheses on neuroimaging data [38,43,91], novel RL agents are proposed to test and refine AI agents in a feedback loop with human experimentation [2], AI agents are tested for flexible behaviour using environments inspired by human experiments [36], or multi-agent deep RL is designed to show comparable hierarchical innovation in problem solving to humans [92].…”

Section: (B) Neuroai Examples In the Rubric Challenges And Opportunitiesmentioning

confidence: 99%

“…What makes an algorithm, representation or architecture neurally plausible or brain-like-be it the brains of Drosophila, rats or humans? Researchers across and within cognitive, neuro-and computer sciences increasingly vary in what they mean when they use terms such as human-like AI [1][2][3], and its close relatives, human-inspired, human-level [4], bio-plausible [5], animal AI [6], as well as brain-inspired [7], neurally plausible, neuroscience-inspired [8] or more recently neuroAI [6,[8][9][10][11][12][13][14][15]. This manuscript proposes that a closer look at research in recent decades reveals three main dimensions of goals and contributions across these fields: a commitment to achieving agents with human-like (or animal-like) behaviour, neural plausibility or to solving specific computer science or engineering problems (mostly irrelevant to neuroscience but perhaps still the goal of brain-like solutions).…”

Section: Introductionmentioning

confidence: 99%

“…Note that these three dimensions are by no means mutually exclusive, binary, nor exhaustive, and the claim is not that all AI researchers should care about neuroAI or the human-likeness of their algorithms. The rubric's goal is to simply provide orientation to varying goals, contributions and intersections, and to clarify how their interactions can advance the fields [1,2,36]. hypotheses to predict or explain the brain's underlying representations and algorithms [17,[37][38][39][40][41][42][43].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

A rubric for human-like agents and NeuroAI

Momennejad

2022

Phil. Trans. R. Soc. B

Self Cite

View full text Add to dashboard Cite

Researchers across cognitive, neuro- and computer sciences increasingly reference ‘human-like’ artificial intelligence and ‘neuroAI’. However, the scope and use of the terms are often inconsistent. Contributed research ranges widely from mimicking behaviour , to testing machine learning methods as neurally plausible hypotheses at the cellular or functional levels, or solving engineering problems. However, it cannot be assumed nor expected that progress on one of these three goals will automatically translate to progress in others. Here, a simple rubric is proposed to clarify the scope of individual contributions, grounded in their commitments to human-like behaviour , neural plausibility or benchmark/engineering/computer science goals. This is clarified using examples of weak and strong neuroAI and human-like agents, and discussing the generative, corroborate and corrective ways in which the three dimensions interact with one another. The author maintains that future progress in artificial intelligence will need strong interactions across the disciplines, with iterative feedback loops and meticulous validity tests—leading to both known and yet-unknown advances that may span decades to come. This article is part of a discussion meeting issue ‘New approaches to 3D vision’.

show abstract

Section: What We Talk About When We Talk About Human-like Behaviourmentioning

confidence: 99%

Section: (E) Human Navigation Turing Testmentioning

confidence: 99%

Section: (B) Neuroai Examples In the Rubric Challenges And Opportunitiesmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

A rubric for human-like agents and NeuroAI

Momennejad

2022

Phil. Trans. R. Soc. B

Self Cite

View full text Add to dashboard Cite

show abstract

Human or AI? The brain knows it! A brain-based Turing Test to discriminate between human and artificial agents.

Pischedda,

Kaufmann,

Wudarczyk

et al. 2023

2023 32nd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN)

View full text Add to dashboard Cite

Navigates Like Me: Understanding How People Evaluate Human-Like AI in Video Games

Milani

Juliani

Momennejad

et al. 2023

Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems

View full text Add to dashboard Cite

We aim to understand how people assess human likeness in navigation produced by people and artificially intelligent (AI) agents in a video game. To this end, we propose a novel AI agent with the goal of generating more human-like behavior. We collect hundreds of crowd-sourced assessments comparing the human-likeness of navigation behavior generated by our agent and baseline AI agents with human-generated behavior. Our proposed agent passes a Turing Test, while the baseline agents do not. By passing a Turing Test, we mean that human judges could not quantitatively distinguish between videos of a person and an AI agent navigating. To understand what people believe constitutes human-like navigation, we extensively analyze the justifications of these assessments. This work provides insights into the characteristics that people consider human-like in the context of goal-directed video game navigation, which is a key step for further improving human interactions with AI agents. CCS CONCEPTS• Applied computing → Computer games; • Computing methodologies → Reinforcement learning; • Human-centered computing → Empirical studies in HCI.

show abstract

How Humans Perceive Human-like Behavior in Video Game Navigation

Cited by 4 publications

References 14 publications

A rubric for human-like agents and NeuroAI

A rubric for human-like agents and NeuroAI

Human or AI? The brain knows it! A brain-based Turing Test to discriminate between human and artificial agents.

Navigates Like Me: Understanding How People Evaluate Human-Like AI in Video Games

Contact Info

Product

Resources

About