12th Dec 2019
How might we...
...help people to gain more control over the offer they’re given
Putting people in control of their content and data
Totes aims to enable the people using the service to be have a positive experience when using the service, supporting them in taking control, eliminating undesired content and creating their own offer preferences.
Product Context
Totes is an online marketplace that sells a wide range of different types of goods from clothes, tech products to video games and occasionally age restricted products. Helping people find what they’re looking for is a key value of the Totes team as well as making precise product recommendations. These recommendations are based on in-app engagement (views or clicks), the historic behavior of other people and purchases made outside the app.
In order to provide the service, Totes is powered by some of the following data:
- People’s previous activities and interest on site
- Purchases made outside the app
- In-app engagement (views or clicks)
Problem & Opportunity
People are often unaware of the inferences about them that drive data-driven services. People also don’t know what they can do to take control of these inferences. Providing transparency into these inferences is an opportunity for Totes to receive feedback on poor recommendations (which is valuable to Totes) while giving people more control over the type of offers they receive.
Design Features
Control of the content
When starting to browse in the marketplace, Totes will inform the person of the possibility to hide undesired content. These choices will feed into the system to create a more tailored offer. The intention is not to let the algorithm create the offer alone, but to create a two way conversation.
Design Features
Understanding more
Hiding undesired content - especially if its a number of times - will stimulate a pop-up question to delve in further. Totes wishes to understand more deeply what is the person’s logic behind deleting a certain type of content? The pop-up offers the opportunity for the person to give explanations on what they’re interested in and why and in turn gives more nuanced input to the AI.
Next steps
In addition to curating people's interests based on the products they personally interact with, Totes could show how other people using the services also influences the AI inferences and ultimately the products shown on Totes. People currently have little or no mental model of how they collectively influence AI.
How might we build on Totes ideas to...
- Build extended profiles on people's preferences in a way that is easy to grasp?
