Kevin - a connected product for security and home automation
For the course "Beautiful Information Products" we created Kevin, an iPad application for machine-learned security and home automation. The assignment was to design a digital product in the context of security, making use of data visualization. We took a bit of a quirky approach and were inspired by a scene in Home Alone, where Kevin McCallister uses Michael Jordan cut-outs and toys to pretend that the whole family is at home.
We originally designed Kevin as an app that can connect with smart devices in your home. By learning about their usage during the day it can replicate your behavior when you’re away from home. Our original idea was that the system would be learning for a specified timeframe, for example a week, to create a clear link between what the system learns and its output.
As a team we came up with the initial concept, afterwards I worked on part of the wireframing. My main responsibility in this project was programming the prototype in Processing. It loads in JSON files with events (e.g. turning on) of devices and uses that to calculate the average time the device is on and the randomness in which it occurs. The data is translated into interface elements in the timeline view and the simulation on the map for a quick overview.
In many connected products like the Nest thermostat, machine learning is used to adapt the product to the needs of the user.
When working properly, an automated system can significantly reduce the mental load on a user. However, when the system fails to understand the user’s intent, it’s hard to grasp based on what information the system makes decisions, and how to overrule these decisions.
Our challenge in this assignment was therefore:
How do you communicate decisions made by an algorithm and how is the user able to override those decisions?
We originally designed Kevin as an app that can connect with smart devices in your home. By learning about their usage during the day it can replicate your behavior when you’re away from home. The system will be learning for a specified timeframe, for example a week, to create a clear link between what the system learns and its output.
We created a timeline view where the data is translated into actionable user-interface elements. The timeline gives an overview of the decisions taken by the system. By dragging the sliders, the user can directly manipulate the schedule and overrule the input of the system. A map-based view of the home can be used to show a quick simulation of how the system will behave when you’re out of the house. With these views, the user is informed but always in control and able to manipulate the smart home from their fingertips.
We intentionally kept the initial proposition of Kevin simple, but while designing and discussing we saw new opportunities emerge. Especially the integration of cameras and sensors of a home security system would enable interesting use cases. First of all, by calibrating the security system and connected products, false security alerts are prevented when devices switch on and off. Secondly, the additional sensors can be used as input for home automation, when a camera picks up that you enter a room it can automatically switch on the lights. This additional level of functionality and dependency between products would be an interesting design challenge to integrate into the existing timeline view.
I hope you enjoyed this project.
How about looking at another one below?