Recycle correctly.

AUG - DEC 2016



reco is a smartwatch app that helps users quickly determine whether or not something is recyclable.

The system intervenes right as you are about to throw something away and guides you to the right bin.

By interacting with reco, users learn through experience about how to properly sort waste.



UX/UI Designer & Researcher

Research Lead, Ideation, Content Development, Prototyping, Interaction Design




PRoblem Space

Waste sorting on GT's campus

We focused on student interactions with source-separated bins around campus.


User research

preliminary interviews

  • 10 novices:  students who regularly consume resources and dispose of waste on campus
  • 4 experts:  representatives of organizations concerned with environmental sustainability at GT

Task-based observations

  • Focus:  learn about influence of knowledge on waste sorting behavior
  • Method:  think-aloud protocol & follow-up interview (5 participants, 3 locations)
  • Procedure:  Participants sorted a list of sample everyday waste items and rated their confidence in each decision.

Natural observations

  • Focus:  learn about the influence of convenience on waste sorting behavior
  • Method:  natural observation (25 students, 4 locations)
  • Procedure:  We observed how students actually dispose of their waste, given the impact of real-world factors such as time pressure, social context, and physical position of bins.

Key Takeaways

1.  We narrowed our problem space from environmental sustainability to waste sorting on GT's campus.

2.  Novices had two main issues when sorting their waste:

Knowledge:  "I'm not sure about all the kinds of materials that are recyclable and which are not - some plastics apparently aren't."

Convenience:  "I definitely care a lot, but convenience gets in the way.  Like with recycling, if it's easier for me to just toss it out I would do that."







Diverge → Converge

Each group member came up with +20 ideas.  We then synthesized and evaluated ideas along 2 dimensions.

User-to-System scale (x-axis)
Ideas on the left (User) are more technologically simple, but require more effort from the user.  Ideas on the right (System) require less user involvement, but require more technology processing.

Transtheoretical Model (y-axis)
We used this model of behavior change technologies.  The model classifies 5 phases in which behavior change can occur (bottom to top): Precontemplation, Contemplation, Preparation, Action, and Maintenance.


top ideas

Passive Lock Screen Training
Users unlock their phone by sorting a piece of virtual trash into a bin. Responses are tracked and data is presented to users so they can learn how to recycle properly.
AR Education Game
Students participate in a campus-wide multiplayer game where they can "capture" territories. This is done by correctly sorting virtual trash into physical bins around campus
Expert Assistance Bin
Students place items into a designated bin when they are confused about whether or not it can be recycled. Experts emotely sort the item through a mobile app and notify the original owner.

Smartwatch App
Proximity sensors on bins will trigger a nudge on users' smartwatches when they are about to throw something away. The app then guides them to the proper bin to dispose of their trash.


final decision

We decided on the smartwatch app based on several criteria.

User Feedback: Based on critique and feedback of the above 4 ideas, the smartwatch app was well-received.
Feasibility: We had to work within our time and resource constraints.
Overall: The smartwatch was an elegant and simple solution.


User interface

We had several whiteboarding sessions to determine how users should interact with the act, given physical, environmental, and other practice constraints.


We created a scenario of how a user might use the app, detailing the journey from start to finish.


Lo-Fi prototype

We created a paper prototype to gather feedback on interacting with a wearable and the concept of the solution.


Users are guided to the correct bin by answering a few yes/no questions about the item they are holding.  We iterated on the information architecture of the decision tree and the wording of the questions.




user flow

reco works by communicating with sensors attached to recycling centers around campus.  As you approach the recycling center, a sensor will trigger when you come within a certain distance of a particular bin.

Each type of bin (Plastic, Landfill, Aluminum, Paper) has it’s own unique proximity sensor.  This proximity sensor will send a notification to your smartwatch.

The initial notification is designed to catch your attention and make you think twice before throwing something away.  To follow up, a short series of simple yes/no questions is used to guide you to the right place to dispose of your trash.


final build

Wizard of Oz

The sensor architecture was difficult to construct given our time limitations.  A Wizard of Oz prototype was built to achieve the effect of proximity sensors.  A wizard dashboard was used to manually trigger a remote notification on the smartwatch.


minimalist UI

To facilitate ease-of-use and minimize distractions:

  1. The interface was designed with a visually simple aesthetic.
  2. We designed basic interactions that can be quickly executed while on-the-go.


The decision tree was refined through multiple iterations, with the following criteria:

  1. Minimal question length - promote readability and comprehension
  2. Minimal tree path length - ensure quick interactions
  3. Robust filtering - system can accurately sort the majority of common waste items

We thank the GT Campus Recycling Coordinator for her support and advice.



The IRB for the following study was recently approved.  User testing will take place in spring 2017.

Proposed study


Assess the efficacy and usability of the reco prototype.


1.  The waste sorting accuracy will be significantly higher for the group using the reco compared to the controlled group.

2.  The overall task time will not be significantly different between the two conditions.

3.  Using reco will lead to an increase in waste sorting knowledge.


40 participants, recruited from Georgia Tech's student body

Between-subjects design

  • Experimental group - complete the waste sorting task while using reco
  • Control group - complete the waste sorting task as normal, without using reco


Pretest survey.  Establish a baseline for participants' existing knowledge of waste sorting and their current behaviors and practices.

Waste sorting task.  Participants will interact with the prototype by completing a waste sorting task.  The task involves sorting predetermined list of sample waste items into a group of bins on Georgia Tech's campus.

Knowledge post-test.  Measure any change in waste sorting knowledge.

Follow-up interview.  The reco group will provide feedback and comments about prototype usability.

Analysis plan

Accuracy - Measured as the percent of waste items correctly sorted.  Compared between groups.

Time - Measured as the amount of time spent sorting each waste item.  Compared between groups.

Knowledge - Measured by the pre/post test comparison.  Compared within groups.

Usability data - Quantitative and qualitative feedback will be used to inform redesign improvements.