Train Chatbot

To create this website, I used React Native, a version of React that is specialised to work on mobile. However, it was still made to work on desktop, too.

The website was created for a University Artificial Intelligence Coursework Project and is for communication with an AI chatbot which recommends the user train tickets based on their requirements. This means the website communicates with a backend made in python using FastAPI.

Features

Usability was very important for this app. The app can be switched between dark mode and light mode with the press of a button, defaulting to the user's system setting.

The app also features a help page, which:

• can be found where in a similar place to other apps, in the top right
• welcomes the user and reassures them that the chatbot will ask for missing inputs
• gives the user an idea of a starting point, reminding them what to ask for
• tells the user of special commands they can input, which are not entirely obvious
• says who made the app

Design

The app's interface was kept purposefully minimal to keep a focus on the bot for both the user and the University project.

The app uses a very simple colour scheme: white, grey and a vibrant blue. The blue was used to highlight important aspects such as buttons and the bot's messages. Making the Help button a vibrant blue is a compromise, making sure the button is visible and visibly separate from the dark mode button but unfortunately perhaps too much attention for users that do not need it.

The app's interface is inspired by Social Media chats like Instagram and WhatsApp.

User Testing

To ensure the app was usable, I ran usability tests with 3 users.

The testers were asked 4 questions, which covered a majority of the possible actions an ordinary user would be able to take. This included making the user purposefully type the wrong name in for a station, so that the bot could try to correct them.

The first user had trouble knowing what to type in, refusing to press the Help button for Help. Therefore, the messages that the bot said had to be updated to try to make the user feel better about pressing the button while also trying to keep them better informed on how to use it without pressing the button.

Fortunately, the next 2 users had absolutely no trouble with the app at all.