Athena AI. Conversational User Interface Design for a Robotaxi… | by Pradhyumnaa G | May 2023

training camp

Conversational UI design for a Robotaxi service.

This story was originally published in my portfolio: Pradhyumnaa G’s Portfolio. I am currently looking for an internship, so if you are interested in my work, please contact me at

The current state of the autonomous car market and Robotaxi.

The state of the self-driving car market
Competitors of the Robotaxi service

What do consumers think of self-driving cars?

Consumer Concerns—Part 1
Consumer Concerns—Part 2
Consumer Concerns—Part 3

We need to build trust between the car and the passenger. Focus on the passenger experience and create a positive first impression

Factors of an intelligent system to build trust
Introducing Athena AI.

We anthropomorphize Athena, so passengers believe she has some semblance of rational thinking and conscious feeling according to this study by the University of Nottingham.

The first attempt at anthropomorphism – The failure
The second attempt at anthropomorphism – The success

Why choose the second option?

Because when anthropomorphizing Athena, we have to keep the strange valley in mind. Also, the first method seems too “static”.

Let’s see how Athena’s conversational UI will work and how it can be implemented.

Possible conversations listed

The vocal prototype

Based on the board, I created a prototype voice to simulate some of the scenarios listed. Click the button below to interact with Athena’s voice prototype. This prototype was created using VoiceFlow.

Click here for the prototype

A potential implementation of this system can be achieved by combining an intention recognition model and a sentiment analysis model. By using a funnel, the model can benefit from separation of concerns and modularity.

Intent recognition model on the Banking77 dataset
Model results on an invisible test set
Sentiment analysis model on a custom dataset

Let’s look at the options using which we can authorize the ride. There are many ways to do this, each with its own advantages and disadvantages and different levels of implementation complexity.

Authorization options compared – QR code is the BEST!

Let’s imagine what a customer’s end-to-end journey would look like when using this service.

Mapping of the customer journey drawn
Embedded User Interface Information Architecture
Wireframes – Horizontal and Vertical Options Compared

Why choose the vertical layout?

I chose to go with the vertical orientation because it’s more suitable for displaying certain types of content, like news articles. There is also a shift in content consumption from desktop to mobile devices.

Click to see the high fidelity prototype

High-fidelity mockup of the on-board user interface
Mobile Application Information Architecture

Click to see the high fidelity prototype

Mobile app high fidelity mockup

Unfortunately, I was unable to contact a large group of participants to perform user testing on the Voice Prototype. However, I was able to gather user feedback through the use of an SUS test for the in-vehicle UI.

In-vehicle UI SUS test result

User comments

From there, the average SUS score of the 13 respondents is 82.5. According to, an SUS score above 68 would be considered above average and anything below 68 is below average. This means that the score of 82.5 is well above average. In addition to this, I received 3 written feedback responses from respondents.

77.5[3]: Seems pretty simple to use and understand, gives clear instructions, is versatile, and provides a lot of things like music and stuff. I think that’s all I need.

90[6]: That’s pretty good for a prototype.

72.5[7]: The user interface seems intuitive and very functional. Although the design may seem friendlier, it is good in the way it looks modern.

Please note that Athena should always be considered an unfinished product. Even after performing the required (extensive) user testing, we will need to continuously evaluate user feedback and be prepared to adjust the interface and chat capabilities to better meet user needs.

Keep in mind that the Intent Recognition and Sentiment Analysis models I developed are just a proof of concept and require extensive research and expertise to create a more robust and accurate system. We will also need to create or collect a high-quality intent dataset to use for Athena’s conversational capabilities to replace the Banking77 proxy dataset.

In addition, the driving and decision-making system of the autonomous vehicle must also be well designed for the safety of passengers and pedestrians. Assuming the Advanced Driver Assistance System (ADAS) is well designed, the Athena can be your value proposition that can differentiate your service from the competition by emphasizing the driving experience it offers to passengers.

If you liked this project and would like to hire me for a UI/UX internship, please contact me at Also, I would like to hear your comments on the project.






Leave a Reply

Your email address will not be published. Required fields are marked *