Engagement Recommendations AI Model

Portfolio > Case Study: Scaling Braindates for E180 > Engagement Recommendations AI Model

Problem

Over several years, the E180 team has developed a set of specialized knowledge on how to deploy braindates to new environments. They have successfully brought braindates to a range of audiences and types of events. To that end, they have learned how to cater to different audiences, introducing them to an entirely new concept and a new technology under a tight timeframe (most events are 2-4 days).

Our challenge was to scale this knowledge, creating a engagement recommendations AI model. This is made form a combination of using metrics for engagement and adoption as well as the internal team's knowledge/intuition around these metrics.

Approach

To begin, I set out to deeply understand what engagement and adoption involves, separating out the metrics that are easily tracked (in the database) with the other datapoints that are taken into consideration when the event team works with clients to develop strategy, but not necessarily tracked systematically.

I performed two major research projects:

  1. User interviews with event team members (to understand engagement challenges)
  2. Designed a working prototype to track 'engagement red-flags'

Interview themes uncovered from experienced event-team members (AKA producers)

This is a Wista Soalbox video for internal event team (producers) to teach them to use the prototype of the machine learning model on engagement recommendations.

 

Solution

I worked in collaboration with a machine learning expert to develop the logic for the model and determine how users would play an active role in helping the model learn. The model creates engagement recommendations to community organizers throughout the process and creates an 'engagement plan' based on the event type, size and audience.

Through this process, we began to understand that users can teach us as much as we can teach them, so interaction with the model is crucial. The idea of ownership and autonomy is important for the users of the product, so we wanted people see and interact with the algorithm actively. We created a system where users rate and feedback into the recommendations to measure their effectiveness. Over time, the company hopes to learn which best practices work in which specific circumstances.

 

Portfolio > Case Study: Scaling Braindates for E180 > Engagement Recommendations AI Model