Streamlining Airport Customer Support

Introducing Gulliver, a Conversational Bot

Airport Information Gathering as a Conversation

Gulliver, or Gull for short, is a low-personification chatbot designed to support travelers throughout their time at the airport with information on available services and concessions, navigation guidance, and helpful reminders.

Open Prototype

Role

Conversation Design, Personality Design, Chatbot Prototyping, User Testing

Team

Bryony H. and Cassandra C.

Duration

4 Weeks

Tools

Voiceflow, Figma

How might a chatbot help take the stress and confusion out of the airport experience?

Transiting through airports and waiting for your flight can be headache-inducing. A friendly chatbot ready to answer questions could make the difference between a positive experience or a stressful one.

We designed our chatbot to be helpful, reliable, and reassuring.

Keeping the context of use in mind, Gull must show care and competence in addressing users’ concerns. Humor and sarcasm would be inappropriate in this context.

Interaction Goals

01

Simplify airport navigation and wayfinding

02

Share updated and relevant information

03

Enhance the airport experience

Personality Design

Gull aims to strike a casual yet competent tone through reassuring language and reliable and helpful information. Recognizing the user's potentially negative emotional state in light of travel snafus, Gull is warm in tone but remains calm in temperament to support the user while acknowledging the reality of their experience.

We compiled our training data to define sample intents, utterances, and slots.

Drawing from our own experiences and common travel issues, we built a robust set of training data including 13 intents and 5-20 utterances for each, a sample of which are below. Within these utterances, we accounted for possible slots such as location, flight number, and other specific keywords.

These intents and utterances were later integrated into our Voiceflow prototype.

We wrote three sample scripts that incorporated our personality design and utterances.

Based on three common scenarios a user might experience, we wrote a sample script for each. They centered around key topics important to travelers while at the airport: dining options, gate location, and boarding.

One of the sample scripts is below, demonstrating how the bot would handle a specific scenario in which the user is seeking vegetarian food options in the airport, as well as pacing and turn-taking. Gull also takes into consideration how the interaction unfolds over space and time, integrating wayfinding through different parts of the airport.

Based on these sample scripts, we diagrammed three conversational flows in Figma.

To visualize the conversational flow, we created a flow diagram for our three sample scripts. Each flow has a unique starting point — an input from the user — but a consistent closing output from Gulliver. The flow diagram also enabled us to determine where error handling may come into play.

What would it be like to actually chat with Gull?

We created a prototype to validate our training data and experience the conversational flows with users.

We used Voiceflow to bring Gulliver to life as a prototyped chatbot.

We created Gull as a custom, code-free AI agent in Voiceflow, ready to automate information sharing related to common airport support questions.

Gull is still in early stages of development. Refer to the scripts and intents to test in the prototype.

Test Prototype

Testers found Gull to be friendly, but not always successful at understanding their request.

We conducted three moderated tests with three separate users and additional testing with ten colleagues. Though the tests with this first-generation prototype were not seamless, users universally praised Gull as "very friendly, inviting, and kind," and said they would use a chatbot like Gull in a real airport setting.

More work needs to be done to strengthen the training data to account for diversity of utterances and better match them with intents to ensure a more fluid conversation.

Where could we go from here?

If we had more time...

Gull was an ambitious project with a lot of potential.

We took on a challenging topic for this project, between the complexity of the airport experience and all its moving parts. The open-ended nature of our conversational flows was tricky for Gull to handle.

If we had more time to fully develop Gull, we might focus on gate-finding or answering more basic questions based on location within the airport.