Casey Kennington
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Casey Kennington has a total of 80 co-authors. Here are the top co-authors by number of shared publications:
- David Schlangen
University of Potsdam
7 shared publications - Jerry Alan Fails
Computer Science, Boise State University, Boise, Idaho, USA
7 shared publications - Maria Soledad Pera
Delft University of Technology, Delft, Netherlands
7 shared publications - Katherine Landau Wright
University of Oregon
6 shared publications - Benjamin Bettencourt
Computer Science, Boise State University, United States
3 shared publications - Catherine Henry
Department of Computer Science Boise State University Boise Idaho 83725
2 shared publications
Latest Publications
- Open Access
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 29-37
Precision or Peril: A PoC of Python Code Quality from Quantized Large Language Models
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 29-37
Open AccessPrecision or Peril: A PoC of Python Code Quality from Quantized Large Language Models
- Open Access
- 19 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Human-Robot Interaction
- Vol. 15 (2) , 1-24
Reporting Guidelines for Large Language Models in Human–Robot Interaction
The comparatively recent advent of Large Language Models (LLMs) has resulted in a wide array of new capabilities and components relevant to Human–Robot Interaction (HRI) researchers. LLMs are being applied to vision, manipulation, planning, reasoning, learning, and HRI problems, frequently as “Scarecrows,” in which LLMs serve as black box modules integrated into robot architectures for the purpose of quickly enabling full-pipeline solutions. However, despite this explosion of applications, general questions remain about the best ways to incorporate LLMs into robot architectures, appropriate safety and guardrail considerations, and, critically, how to report properly on HRI research that involves LLMs. In this article, we explore the question of reporting guidelines for HRI researchers who utilize Scarecrows in robot architectures. We identify five key stakeholder groups in the HRI research process, discuss what information each group needs from HRI researchers, and identify appropriate mechanisms for conveying that information from HRI researchers to stakeholders either directly or indirectly. We contribute a set of suggested guidelines regarding what information should be included when researchers disseminate information about HRI research that uses LLMs.
- 19 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Human-Robot Interaction
- Vol. 15 (2) , 1-24
Open AccessReporting Guidelines for Large Language Models in Human–Robot Interaction
The comparatively recent advent of Large Language Models (LLMs) has resulted in a wide array of new capabilities and components relevant to Human–Robot Interaction (HRI) researchers. LLMs are being applied to vision, manipulation, planning, reasoning, learning, and HRI problems, frequently as “Scarecrows,” in which LLMs serve as black box modules integrated into robot architectures for the purpose of quickly enabling full-pipeline solutions. However, despite this explosion of applications, general questions remain about the best ways to incorporate LLMs into robot architectures, appropriate safety and guardrail considerations, and, critically, how to report properly on HRI research that involves LLMs. In this article, we explore the question of reporting guidelines for HRI researchers who utilize Scarecrows in robot architectures. We identify five key stakeholder groups in the HRI research process, discuss what information each group needs from HRI researchers, and identify appropriate mechanisms for conveying that information from HRI researchers to stakeholders either directly or indirectly. We contribute a set of suggested guidelines regarding what information should be included when researchers disseminate information about HRI research that uses LLMs.
- Open Access
- 15 December 2025
- Published by University of Illinois Libraries in Dialogue & Discourse
- Vol. 16 (3) , 96-130
Prior Lessons of Incremental Dialogue and Robot Action Management for the Age of Language Models
Efforts towards endowing robots with the ability to speak have benefited from recent advancements in natural language processing, in particular large language models. However, current language models are not fully incremental, as their processing is inherently monotonic and thus lack the ability to revise their interpretations or output in light of newer observations. This monotonicity has important implications for the development of dialogue systems for human–robot interaction. In this paper, we review the literature on interactive systems that operate incrementally (i.e., at the word level or below it). We motivate the need for incremental systems, survey incremental modeling of important aspects of dialogue like speech recognition and language generation. Primary focus is on the part of the system that makes decisions, known as the dialogue manager. We find that there is very little research on incremental dialogue management, offer some requirements for practical incremental dialogue management, and implications of incremental dialogue for embodied, robotic platforms in the age of large language models.
- 15 December 2025
- Published by University of Illinois Libraries in Dialogue & Discourse
- Vol. 16 (3) , 96-130
Open AccessPrior Lessons of Incremental Dialogue and Robot Action Management for the Age of Language Models
Efforts towards endowing robots with the ability to speak have benefited from recent advancements in natural language processing, in particular large language models. However, current language models are not fully incremental, as their processing is inherently monotonic and thus lack the ability to revise their interpretations or output in light of newer observations. This monotonicity has important implications for the development of dialogue systems for human–robot interaction. In this paper, we review the literature on interactive systems that operate incrementally (i.e., at the word level or below it). We motivate the need for incremental systems, survey incremental modeling of important aspects of dialogue like speech recognition and language generation. Primary focus is on the part of the system that makes decisions, known as the dialogue manager. We find that there is very little research on incremental dialogue management, offer some requirements for practical incremental dialogue management, and implications of incremental dialogue for embodied, robotic platforms in the age of large language models.
- Open Access
- 14 November 2025
- Published by Frontiers Media SA in Frontiers in Robotics and AI
- Vol. 12, 1725162
Editorial: Dialogue with robots: constructive approaches for understanding communication
In Groß et al. (2023a), the authors present RISE, or Robotics Integration and Scenario-Management Extensible-Architecture, an open-source framework for reproducible HRI research. RISE addresses challenges in HRI research by offering an accessible and configurable system built on three core structures: (1) communication robot acts, (2) interaction rules, and (3) working memory. RISE offers HRI experiment designers an easy-to-use graphical user interface and bindings with ROS middleware.The field of social robotics puts focus on the social impacts of robots (as opposed to, for example, how to build a functioning robotic arm). Huang and Moore ( 2023) is an exploratory study on how a social robot's affordances (its appearance, voice, and behaviors) affect user perceptions in conversational HRI. The researchers found that initial warmth is significantly influenced by static affordances, with child-like robots starting strongly but declining post-interaction. The authors highlight the necessity of aligning affordances (both static and dynamic) with the robot's intended role and ensuring genuine, responsive interaction to manage user expectations and bridge the "habitability gap" which refers to the discrepancy between a robot's capabilities and what a user might expect the robot should be able to do.In a similar line of research, Robb et al. (2023) considered how robot morphology affects conversational interaction. The findings indicate that the robot, with its anthropomorphic features and social cues, generally fostered higher engagement and was trusted more in the high-stakes emergency context (in their experiment, an off-shore energy platform with a time constraint to resolve the issue) compared to the voice-only smart speaker. Their work shows that embodiment could be a key factor in successfully deploying conversational agents in the professional workplace.In Figueroa et al. (2023), the authors highlight the growing social issue of loneliness and the burden on healthcare systems, suggesting that social robots could be a promising solution. Key findings from interviews with the participants from the memory clinic and interaction logs indicate that participants gradually accepted the robot, developed a sense of attachment and companionship, and maintained regular usage over several months, suggesting a positive influence on their daily lives.Continuing on the theme of social robotics, Lumer & Buschmeier (2023) explore how expectations of robot politeness can affect interactions between humans and robots. The authors identified two types of politeness: adaptive (i.e., politeness oriented toward an individual listener) and rule-governed (i.e., politeness that follows cultural and societal norms). Their findings indicate that while humans use both types of politeness, users primarily expect robots to exhibit only the functional, rule-governed politeness due to the perceived lack of "feelings" or agency in the artificial agents. This distinction offers valuable design implications for enhancing the user experience of social robots, suggesting that rule-governed politeness is suitable for public settings, while adaptive, customized politeness may be desired in private settings.When robots interact with humans using spoken dialogue, humans bring expectations that robots have a degree of emotional awareness in the interactions. Mishra et al. (2023) investigates the use of Large Language Models (LLMs) for real-time emotion generation in human-robot dialogue. This work highlights the potential of LLMs to move beyond just generating speech by controlling the affective-emotional behavior of robots in real-time, for applications in areas like companionship and customer service.Recent research is leading to human-robot communication becoming more natural, but there are some important considerations that need to be first addressed. Groß et al (2023b) details an empirical study investigating the effectiveness of using negation as a contrastive guidance strategy within explanatory human-robot dialogue for task performance. The study concludes that incorporating negation can enhance the naturalness and effectiveness of robot-to-human explanations, supporting the goal of developing more adaptive and communicative artificial intelligence.In a similar line of research, Kawakubo et al. (2024) explored asymmetries in human-robot communication. The authors define asymmetry as the situation where a robot treats a human like a person, but the human treats the robot like a machine (perhaps surprisingly due to the highly anthropomorphic robots used by the authors). Experiments using simulated dialogue videos indicated that this "pretending to tailor" strategy, while non-human-like, could positively enhance the perception of the robot's effort to adapt, particularly among customers who already view the robot as a system, thereby proving effective for a human-robot symbiotic society.Siskind et al. ( 2024) begins with a quote from Arthur C. Clarke: "Any sufficiently advanced technology is indistinguishable from magic" upon which the authors elaborate: "A magician's trick and a chatbot conversation have something in common: most of their audiences do not know how they work." The researchers identify several key psychological techniques from magic, such as misdirection, controlling expectations, and emotional connection, and adapt them to a social robot named Haru. The study advocates for leveraging the showmanship and psychological mastery of magicians to create more impressive, satisfying, and memorable interactions with conversational AI.Li & Ross (2023) present a controlled HRI study aimed at systematically invoking and identifying different states of user confusion during task-oriented dialogues. The authors used a Wizard-of-Oz design with a Pepper robot to trigger non-confusion, productive confusion (i.e., when a human user recognizes and attempts to directly resolve the confusion), and unproductive...
- 14 November 2025
- Published by Frontiers Media SA in Frontiers in Robotics and AI
- Vol. 12, 1725162
Open AccessEditorial: Dialogue with robots: constructive approaches for understanding communication
In Groß et al. (2023a), the authors present RISE, or Robotics Integration and Scenario-Management Extensible-Architecture, an open-source framework for reproducible HRI research. RISE addresses challenges in HRI research by offering an accessible and configurable system built on three core structures: (1) communication robot acts, (2) interaction rules, and (3) working memory. RISE offers HRI experiment designers an easy-to-use graphical user interface and bindings with ROS middleware.The field of social robotics puts focus on the social impacts of robots (as opposed to, for example, how to build a functioning robotic arm). Huang and Moore ( 2023) is an exploratory study on how a social robot's affordances (its appearance, voice, and behaviors) affect user perceptions in conversational HRI. The researchers found that initial warmth is significantly influenced by static affordances, with child-like robots starting strongly but declining post-interaction. The authors highlight the necessity of aligning affordances (both static and dynamic) with the robot's intended role and ensuring genuine, responsive interaction to manage user expectations and bridge the "habitability gap" which refers to the discrepancy between a robot's capabilities and what a user might expect the robot should be able to do.In a similar line of research, Robb et al. (2023) considered how robot morphology affects conversational interaction. The findings indicate that the robot, with its anthropomorphic features and social cues, generally fostered higher engagement and was trusted more in the high-stakes emergency context (in their experiment, an off-shore energy platform with a time constraint to resolve the issue) compared to the voice-only smart speaker. Their work shows that embodiment could be a key factor in successfully deploying conversational agents in the professional workplace.In Figueroa et al. (2023), the authors highlight the growing social issue of loneliness and the burden on healthcare systems, suggesting that social robots could be a promising solution. Key findings from interviews with the participants from the memory clinic and interaction logs indicate that participants gradually accepted the robot, developed a sense of attachment and companionship, and maintained regular usage over several months, suggesting a positive influence on their daily lives.Continuing on the theme of social robotics, Lumer & Buschmeier (2023) explore how expectations of robot politeness can affect interactions between humans and robots. The authors identified two types of politeness: adaptive (i.e., politeness oriented toward an individual listener) and rule-governed (i.e., politeness that follows cultural and societal norms). Their findings indicate that while humans use both types of politeness, users primarily expect robots to exhibit only the functional, rule-governed politeness due to the perceived lack of "feelings" or agency in the artificial agents. This distinction offers valuable design implications for enhancing the user experience of social robots, suggesting that rule-governed politeness is suitable for public settings, while adaptive, customized politeness may be desired in private settings.When robots interact with humans using spoken dialogue, humans bring expectations that robots have a degree of emotional awareness in the interactions. Mishra et al. (2023) investigates the use of Large Language Models (LLMs) for real-time emotion generation in human-robot dialogue. This work highlights the potential of LLMs to move beyond just generating speech by controlling the affective-emotional behavior of robots in real-time, for applications in areas like companionship and customer service.Recent research is leading to human-robot communication becoming more natural, but there are some important considerations that need to be first addressed. Groß et al (2023b) details an empirical study investigating the effectiveness of using negation as a contrastive guidance strategy within explanatory human-robot dialogue for task performance. The study concludes that incorporating negation can enhance the naturalness and effectiveness of robot-to-human explanations, supporting the goal of developing more adaptive and communicative artificial intelligence.In a similar line of research, Kawakubo et al. (2024) explored asymmetries in human-robot communication. The authors define asymmetry as the situation where a robot treats a human like a person, but the human treats the robot like a machine (perhaps surprisingly due to the highly anthropomorphic robots used by the authors). Experiments using simulated dialogue videos indicated that this "pretending to tailor" strategy, while non-human-like, could positively enhance the perception of the robot's effort to adapt, particularly among customers who already view the robot as a system, thereby proving effective for a human-robot symbiotic society.Siskind et al. ( 2024) begins with a quote from Arthur C. Clarke: "Any sufficiently advanced technology is indistinguishable from magic" upon which the authors elaborate: "A magician's trick and a chatbot conversation have something in common: most of their audiences do not know how they work." The researchers identify several key psychological techniques from magic, such as misdirection, controlling expectations, and emotional connection, and adapt them to a social robot named Haru. The study advocates for leveraging the showmanship and psychological mastery of magicians to create more impressive, satisfying, and memorable interactions with conversational AI.Li & Ross (2023) present a controlled HRI study aimed at systematically invoking and identifying different states of user confusion during task-oriented dialogues. The authors used a Wizard-of-Oz design with a Pepper robot to trigger non-confusion, productive confusion (i.e., when a human user recognizes and attempts to directly resolve the confusion), and unproductive...
- Open Access
- 10 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 575-576
Retico: A Framework for Robot/IVA-ready Spoken Dialogue
- 10 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 575-576
Open AccessRetico: A Framework for Robot/IVA-ready Spoken Dialogue
- Open Access
Could the Road to Grounded, Neuro-symbolic AI be Paved with Words-as-Classifiers?
Formal, Distributional, and Grounded theories of computational semantics each have their uses and their drawbacks. There has been a shift to ground models of language by adding visual knowledge, and there has been a call to enrich models of language with symbolic methods to gain the benefits from formal, distributional, and grounded theories. In this paper, we attempt to make the case that one potential path forward in unifying all three semantic fields is paved with the words-as-classifier model, a model of word-level grounded semantics that has been incorporated into formalisms and distributional language models in the literature, and it has been well-tested within interactive dialogue settings. We review that literature, motivate the words-as-classifiers model with an appeal to recent work in cognitive science, and describe a small experiment. Finally, we sketch a model of semantics unified through words-as-classifiers.
Open AccessCould the Road to Grounded, Neuro-symbolic AI be Paved with Words-as-Classifiers?
Formal, Distributional, and Grounded theories of computational semantics each have their uses and their drawbacks. There has been a shift to ground models of language by adding visual knowledge, and there has been a call to enrich models of language with symbolic methods to gain the benefits from formal, distributional, and grounded theories. In this paper, we attempt to make the case that one potential path forward in unifying all three semantic fields is paved with the words-as-classifier model, a model of word-level grounded semantics that has been incorporated into formalisms and distributional language models in the literature, and it has been well-tested within interactive dialogue settings. We review that literature, motivate the words-as-classifiers model with an appeal to recent work in cognitive science, and describe a small experiment. Finally, we sketch a model of semantics unified through words-as-classifiers.
- Open Access
- 17 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 62-69
How Readability Cues Affect Children's Navigation of Search Engine Result Pages
Children often interact with search engines within a classroom context to complete assignments or discover new information. To successfully identify relevant resources among those presented on a search engine results page (SERP), users must first be able to comprehend the text included in SERP snippets. While this task may be straightforward for an adult user, children may encounter obstacles in terms of readability and comprehension when attempting to navigate a SERP. Previous research has demonstrated the positive impact of including visual cues on a SERP as relevance signals to guide children toward appropriate resources. In this work, we explore the effect of supplying visual cues related to readability and text difficulty on children’s (ages 6-12) navigation of a SERP. Using quantitative data collected from user-interface interactions and qualitative data gathered from participant interviews, we analyze the impact of these visual cues on children’s selection of results on a SERP when carrying out information discovery tasks.
- 17 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 62-69
Open AccessHow Readability Cues Affect Children's Navigation of Search Engine Result Pages
Children often interact with search engines within a classroom context to complete assignments or discover new information. To successfully identify relevant resources among those presented on a search engine results page (SERP), users must first be able to comprehend the text included in SERP snippets. While this task may be straightforward for an adult user, children may encounter obstacles in terms of readability and comprehension when attempting to navigate a SERP. Previous research has demonstrated the positive impact of including visual cues on a SERP as relevance signals to guide children toward appropriate resources. In this work, we explore the effect of supplying visual cues related to readability and text difficulty on children’s (ages 6-12) navigation of a SERP. Using quantitative data collected from user-interface interactions and qualitative data gathered from participant interviews, we analyze the impact of these visual cues on children’s selection of results on a SERP when carrying out information discovery tasks.
- Open Access
- 17 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 828-833
Kid Query: Co-designing an Application to Scaffold Query Formulation
In this work, we discuss the findings emerging from co-design sessions between children ages 6 to 11 and adults, which were conducted to advance knowledge on how to best support children using well-known search tools for online information discovery. Specifically, we argue that by leveraging scaffolding, gamification techniques, and design choices via an application, it is possible to enhance children’s habits related to query formulation. Outcomes from this preliminary exploration reveal that gameplay incentives (e.g. levels, points, and other incentives like customization) are needed and effective in motivating further interaction with the application, which in turn leads to further utilization of the scaffolding needed to positively impact query formulation.
- 17 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 828-833
Open AccessKid Query: Co-designing an Application to Scaffold Query Formulation
In this work, we discuss the findings emerging from co-design sessions between children ages 6 to 11 and adults, which were conducted to advance knowledge on how to best support children using well-known search tools for online information discovery. Specifically, we argue that by leveraging scaffolding, gamification techniques, and design choices via an application, it is possible to enhance children’s habits related to query formulation. Outcomes from this preliminary exploration reveal that gameplay incentives (e.g. levels, points, and other incentives like customization) are needed and effective in motivating further interaction with the application, which in turn leads to further utilization of the scaffolding needed to positively impact query formulation.
- Open Access
- 1 May 2024
- Published by Boise State University, Albertsons Library in BOGA: Basque Studies Consortium Journal
- Vol. 11, 2
Word Concreteness and Word Originality
Basque is the oldest known spoken language of Western Europe that is not descended from Indo-European. Despite efforts to stop its use from external influences, the Basque people have shown resiliency throughout their history to keep the language alive. However, as is the case with all other languages, Basque has not been immune to influence from neighboring languages; indeed all Basque speakers are bilingual. The question explored here is whether a word’s concreteness using machine translation from Basque to English and English concreteness scores determines if an original Basque word is spoken today. We find that there is a significant difference of concreteness scores between old and contemporary Basque. We also compared old Basque words with common contemporary Basque words in a semantic vector space and found the old words to cluster more tightly than contemporary words suggesting that words with similar semantic vectors are able to retain usage through time.
- 1 May 2024
- Published by Boise State University, Albertsons Library in BOGA: Basque Studies Consortium Journal
- Vol. 11, 2
Open AccessWord Concreteness and Word Originality
Basque is the oldest known spoken language of Western Europe that is not descended from Indo-European. Despite efforts to stop its use from external influences, the Basque people have shown resiliency throughout their history to keep the language alive. However, as is the case with all other languages, Basque has not been immune to influence from neighboring languages; indeed all Basque speakers are bilingual. The question explored here is whether a word’s concreteness using machine translation from Basque to English and English concreteness scores determines if an original Basque word is spoken today. We find that there is a significant difference of concreteness scores between old and contemporary Basque. We also compared old Basque words with common contemporary Basque words in a semantic vector space and found the old words to cluster more tightly than contemporary words suggesting that words with similar semantic vectors are able to retain usage through time.
- Open Access
Unsupervised, Bottom-up Category Discovery for Symbol Grounding with a Curious Robot
Towards addressing the Symbol Grounding Problem and motivated by early childhood language development, we leverage a robot which has been equipped with an approximate model of curiosity with particular focus on bottom-up building of unsupervised categories grounded in the physical world. That is, rather than starting with a top-down symbol (e.g., a word referring to an object) and providing meaning through the application of predetermined samples, the robot autonomously and gradually breaks up its exploration space into a series of increasingly specific unlabeled categories at which point an external expert may optionally provide a symbol association. We extend prior work by using a robot that can observe the visual world, introducing a higher dimensional sensory space, and using a more generalizable method of category building. Our experiments show that the robot learns categories based on actions and what it visually observes, and that those categories can be symbolically grounded into.https://info.arxiv.org/help/prep#comments
Open AccessUnsupervised, Bottom-up Category Discovery for Symbol Grounding with a Curious Robot
Towards addressing the Symbol Grounding Problem and motivated by early childhood language development, we leverage a robot which has been equipped with an approximate model of curiosity with particular focus on bottom-up building of unsupervised categories grounded in the physical world. That is, rather than starting with a top-down symbol (e.g., a word referring to an object) and providing meaning through the application of predetermined samples, the robot autonomously and gradually breaks up its exploration space into a series of increasingly specific unlabeled categories at which point an external expert may optionally provide a symbol association. We extend prior work by using a robot that can observe the visual world, introducing a higher dimensional sensory space, and using a more generalizable method of category building. Our experiments show that the robot learns categories based on actions and what it visually observes, and that those categories can be symbolically grounded into.https://info.arxiv.org/help/prep#comments