Institute of Art Design + Technology
Dún Laoghaire
On Show
 Courses
Search
Close
3D Animation 3D Design Modelmaking + Digital Art Animation Applied Psychology Art Character MakeUp Design Costume Design Creative Computing Creative Music Production Cyberpsychology Design for Change Film + Television Production Interaction + User Experience Design New Media Studies Photography Production Design User Experience Design Visual Communication Design




Fernanda Alves 

MSc User Experience Design

Hi, I'm a UX designer with a background in visual arts, graphic design, and front-end development. I love making technology better for people. To me, it's essential that products and services are developed with users in mind to create truly meaningful experiences. With the rise of technologies like generative AI, we need to find new ways to enhance this connection. Curiosity, empathy, and problem-solving are at the heart of UX design, but it’s crucial to make sure we’re solving the right problems. That’s where my analytical skills come in, helping to pinpoint and tackle the real issues effectively.

This project explores how onboarding screens in AI chatbots, based on explainable AI concepts, affect user trust, acceptance, and usability. My interest in this area started with my curiosity about ChatGPT.
Artificial intelligence (AI) mimics human thinking and behaviour. AI chatbots, often found in messaging apps and websites, understand and respond to users in natural language. Machine learning (ML), a part of AI, uses algorithms to help software learn from data. This learning happens through large, pre-trained models and deep learning, where machines find patterns in raw data using neural networks. Generative AI is a type of AI that creates new content like text, images, audio, and video using these techniques. However, the complex nature of these models creates a "black box" effect, where the decision-making process isn't clear to humans. Explainable AI (XAI) aims to make these processes understandable, building trust and comprehension in AI outcomes. The gap lies in making these advanced AI technologies more transparent to users, so people can better understand and trust the systems they interact with.
The research questions and hypotheses were created based on the gaps found in the literature review, focusing on trust, acceptance of technology, and usability.
This research adopted the  Stanford Design Thinking Process, which is composed of five steps: 1.    Empathise. 2.    Define. 3.    Ideate.  4.    Prototype.  5.    Test.
The user Erin was created based on insights and pain points demonstrated in surveys and interviews.
The user’s journey describes Erin’s experience while mapping the opportunities for improvement.
The paper prototype is part of the ideation phase. It is the result of the first opportunities identified in previous steps, such as how to include contextual menus and information on the AI chatbot. The paper prototype facilitated the development of the digital prototype.
Version without onboarding
Version with onboarding
The experiment results were analysed through hypothesis testing, and the statistical methodology utilised was the t-test: two-sample assuming two equal variances. According to the results, none of the hypotheses could be statistically proven.
The qualitative results showed that users had a more positive perception of the AI chatbot when onboarding screens were included. Participants more frequently rated messages as "accurate" when using the onboarding version. Additionally, users successfully completed more tasks with the onboarding version compared to the version without it.
This project explored how onboarding features could help users calibrate their trust in AI chatbots. Using both quantitative and qualitative methods, it examined hypotheses on aspects such as trust, acceptance, and usability. Qualitative findings indicated that onboarding could enhance users' understanding and trust in AI chatbots, but more research is required to identify specific aspects.
Project Description

This research focused on enhancing the transparency and trustworthiness of AI systems, particularly chatbots like ChatGPT. Explainable AI (XAI) is an emerging field dedicated to helping users understand and trust AI decisions.
However, there is still much to learn about how to effectively communicate these AI processes to users.
To address this gap, this project investigated the use of onboarding processes to improve trust, acceptance, and perceived usability of AI chatbots.


Onboarding in AI Chatbots: Explaining for Trust

This research examined how onboarding screens in AI chatbots, based on explainable AI concepts, affect user trust, satisfaction, and usability. Through an A/B test, two versions of a chatbot were created—one with onboarding and one without. Although no statistical evidence supported the hypotheses, qualitative findings highlighted some positive effects of onboarding, such as improved accuracy, perception, performance, and usability. These results suggest that effective onboarding should also consider the complexity and timing of explanations and users' expertise. Future research should explore which specific onboarding aspects are most beneficial.

User Experience Design
Top