Portfolio

At its core lies a deep dive into the fundamental concept of user expectations of social robots and their AI systems, drawing insights from psychological theories, UX methodology and real-world scenarios to illuminate their profound impact on social encounters. Through meticulous examination, I asked questions and shed light on the unique challenges posed by social robots — sophisticated machines that blur the line between human-like interaction partners and inanimate objects. My doctoral thesis represents a significant contribution to the field of human-robot interaction, offering valuable insights into the complexities of user expectations and practical guidelines for managing them in research and in society.

Key findings

• Complexity of User Expectations: User expectations in HRI are influenced by a myriad of factors, ranging from past experiences to societal norms. Social robots, with their dual nature as human-like entities and machines, present unique challenges in managing user expectations.
• Social Robot Expectation Gap: The disparity between users’ expectations of social robots and their actual capabilities—termed the social robot expectation gap—poses a significant challenge in HRI. This gap can lead to misunderstandings and dissatisfaction in human-robot interactions.
• Guidelines for Managing Expectations: As part of my research, I have developed practical guidelines for managing user expectations in human-robot interaction. These guidelines provide insights for researchers and designers to bridge the expectation gap and enhance user experience with social robots for better integration in society.
• Transparency and Communication: Transparent communication about social robots, their capabilities, and limitations is essential for managing user expectations effectively. By fostering clear and honest dissemination of information, we can mitigate the social robot expectation gap and foster more positive interactions between humans and robots.
• Societal Integration: Investigated strategies and approaches for the integration of social robots andtheir AI systems into society, emphasizing the importance of managing user expectations to ensure smooth integration and acceptance in various social contexts.

Outcome

The doctoral thesis was published and defended in January 2024, titled “What did you expect? A human-centered approach to investigating and reducing the social robot expectation gap”.

Key Features

• Conducted user tests to optimize dialogue system parameters.
• Evaluated the overall performance and user experience of the dialogue system.
• Integrated GPT-3 with Aldebaran Pepper and Nao robots for verbal interactions.
• Collaborated with a multidisciplinary team to implement the system.

Technical Description

• Implemented four software components, including a chatbot service, chatbot bridge, speech recognition module, and dialogue module.
• Utilized NaoQi middleware, Google Cloud Speech-to-Text, and OpenAI GPT-3 API for system functionality.
• Integrated Python 2.7 and Python 3.10 components using ZeroMQ to bridge communication.
• Developed mechanisms for audio capture, speech-to-text conversion, and robot speech generation.

Outcome

The system was showcased live at the the 18th ACM/IEEE International Conference on Human-Robot Interaction (HRI 2023), offering participants the opportunity to engage in real-time dialogue with the robots, and is published in the Companion for the Conference. The source code was made publicly available to foster community collaboration in designing and evaluating new dialogue systems for robots.

This framework offers a structured approach to comprehensively study and manage user expectations in sHRI, thereby enhancing overall user experience and informing the design of future robotic systems. It serves as a cornerstone for empirical studies in this domain, offering a comprehensive toolset to explore and address the intricate interplay of user expectations and robotic interactions.

Key Findings

• Identified three primary factors influencing expectations: affect, cognitive processing, and behavior & performance.
• Introduced key metrics for evaluation including The Negative Attitudes Towards Robots Scale (NARS), The Robot Anxiety Scale (RAS), questions related to Closeness, question related to Perceived Capability, memory recall, reaction time, choice of dialogue, word repetition, interruptions, and interaction duration.
• Formulated four UX goals to guide interaction design and evaluation.
• Outlined a systematic procedure for studying expectations, involving scenario identification, data collection, analysis, and reporting.

Outcome

I presented this theoretical work at the 24th International Conference on Human-Computer Interaction and first-authored the paper published by Springer International in the conference proceedings, titled “The Social Robot Expectation Gap Evaluation Framework” (2022). Additionally, this work is part of doctoral thesis.

The findings highlight the importance of considering users’ previous experiences with robots when designing and deploying social robots. Understanding the factors that shape users’ expectations is crucial for enhancing the overall user experience and acceptance of robotic systems. The findings contribute to the growing body of knowledge in human-robot interaction research and inform the design of future robotic systems aimed at delivering satisfying user experiences.

Data Collection and Analysis

The experiment employed a within-subject design, utilizing the Social Robot Expectation Gap Evaluation Framework, measuring expectations using subjective measures. Data was collected before, during, and after interactions with Pepper, and hypotheses were formulated to test the influence of time and previous experience on participants’ expectations. Statistical analyses including two-sided F-tests and repeated measures ANOVA were performed to analyze the data.

Key Findings

• Variability in participants’ expectations did not significantly decrease over time, indicating robust and persistent expectations.
• Previous experience with robots significantly influenced participants’ responses, suggesting a strong impact on expectations.
• Participants’ affect towards the robot changed in both positive and negative directions over the course of interactions, indicating dynamic shifts in expectations.
• While some measures changed over time, others remained stable, reflecting nuanced changes in expectations.

Outcome

The results of this work are presented in a publication, which I first-authored, in The International Journal of Social Robotics, titled “Previous Experience Matters: An In-Person Investigation of Expectations in Human-Robot Interaction” (2024). Additionally, this work is part of my doctoral thesis.

Using the Social Robot Expectation Gap Evaluation Framework, it assessed four UX goals based on observations, interviews, and subjective measures. Participants interacted with the social robot Pepper twice in order to evaluate how expectations change over time. I designed and conducted the study, developed research protocols, recruited participants, facilitated interactions, collected and analyzed data, and drew conclusions.

Key Findings

• Initial interactions showed anxious behavior and uncertainty, indicating a lack of experience.
• Participants improved in cognitive processing over time but struggled with interaction uncertainty.
• Communication issues, one-sided interactions, and speech recognition problems hindered engagement.
• Participants felt interrupted, with conversations affected by uncertainty about the robot’s capabilities.

Outcome

I presented the findings at the 25th International Conference on Human-Computer Interaction and first-authored the paper published by Springer International in the conference proceedings, titled “Applying the Social Robot Expectation Gap Evaluation Framework” (2023). Additionally, this work is part of my doctoral thesis.

Data Collection and Analysis

The qualitative analysis, utilizing the Social Robot Expectation Gap Evaluation Framework, drew data from post-test interviews, video recordings, and field notes collected from all 31 participants. A reflexive thematic analysis approach was employed, consisting of six phases: familiarization with the data, initial code generation, theme generation, theme review, theme definition, and report writing. The analysis focused on understanding the interaction quality, identifying interaction strategies, exploring confirmed and disconfirmed expectations, examining core elements impacting participants, and understanding users’ experiences.

Key Findings

• Participants employed various interaction strategies, including getting to know the robot, exploring its capabilities, and testing its limits.
• Expectations varied across a spectrum, with participants demonstrating adaptability based on their previous experiences and encounters.
• Interaction quality improved from the first to the second session, indicating participants’ heightened understanding of how to engage with the robot.
• Participants exhibited varied reactions to different aspects of the robot, highlighting surprises and concerns regarding its capabilities and human-likeness.
• Users’ expectations and experiences were influenced by factors such as dialogue quality, response timing, and the robot’s appearance.

Outcome

The results of this work is currently submitted for scientific journal. Additionally, this work is part of my doctoral thesis.