CASE STUDY: "SPARK" - A.I. PROTOTYPE


We designed an Artificial Intelligence based application that sparks conversation.


Project: Hack-a-thon 2018

Time: June 2018


Role: Creative Director, UX Lead, Presented Ideas


THE ASK: Develop a consumer experience that demonstrates to attendees the power of A.I. through a hospitality experience and/or a consumer experience.


THE SOLUTION: see pdf here


Defining Values:

Engaging conversations build meaningful relationships between individuals everyday.

We believe A.I. has the potential to bring us closer and lead to new human-machine interfaces.

Our Objective:

Create an A.I. solution that enhances experiences as well as generates an increasing understanding of human conversations.

Solution:

Spark is an intelligent assistant that uses guest/consumer data to produce insightful discussion topics. The A.I. platform will learn through usage in order to create better future machine – human interfaces and interactions.

Spark facilitates conversation by finding common ground between strangers. It suggests best-fit topics that pertain to everyone within the immediate local setting – i.e. Dinner Table, Kiosk, Booth, Office, etc. Or it finds a common tangentially-related topic that spark a conversation and connection.


Spark does this by aggregating data from social platforms, music streaming services, GMR’s Orchestrate app and/or digital questionnaires.


This A.I. platform is a software application that lives in the cloud and will be brought to life on your phone via an app or through existing digital assistants such as Google Home or Alexa.

These conversation starters are generated in real-time and use shared interested between people. This usage helps Spark to learn via the Hack.ai platform so it gets better results the more it is used. And, of course, it translates for each person at the table.


Utilizing Google’s TensorFlow for Machine Learning, we will analyze topics and categories to figure out what people really like to open up about.

Our Label will be whether a given group of participants will enjoy a conversation topic. Our Features will include:

•Acceptance of the Topic – Yes/No

•Length of the discussion in seconds

•Desire to continue the discussion with a new topic – Yes/No

•Decided by the acceptance flag on the next offer

•New social media connections with other participants

• For the purposes of this demonstration, users will not have accounts or be required to log in. We will simply detect participants when they invoke the application on a smart speaker.