Human-Chatbot Interaction: Assessing confidence, technology acceptance & trust in chatbots & AI agents
City St. George, University of London, UK
2018 – 2020
This paper is currently being converted into a summarized, presentable deck to be released in the next few months.
Overview
Over the years, there has been an increasing interest in automated technologies, including chatbots that rely on advancements in artificial intelligence (AI) and machine learning. Nowadays, chatbots are used in many application areas. Various reports (Deloitte, 2017; Gartner, 2018; Juniper, 2018) estimated a usage increase by up to 85% by 2020 and 2023. Nonetheless, a knowledge gap concerning the intent to use and trust in chatbots still exists.
This paper highlights factors influencing acceptance, trust, and confidence in chatbots. This was investigated by administering 2 chatbots, using different technologies and deployed in different application areas, to a sample of 14 participants (n=14) (7 females, 7 males). Data was collected through interviews, observations, and a human-computer trust survey validated by previous research.
Methods
- Within-subjects design experiment
- Semi-structured interviews
- Moderated observations (with retrospective think-alouds)
- Thematic analysis (deductive approach)
- Survey study (human-computer trust)
- Paired-samples t-Test with Bonferroni Adjusted Alpha applied
Outcomes
Findings indicated that the application area predicts chatbot usage. This is influenced by other factors like understandability, reliability, usefulness, risks, and external factors (e.g., Cambridge Analytica). It was noticed that human likeness induced negative emotions among female participants, which led to the ‘uncanny valley effect’. In contrast, males were found to have high trust propensity in chatbots. Such discoveries suggest that these factors, combined with gender differences and attitudes, could predict trust and chatbot acceptance.
Paper
More case studies
