Context is the problem and the solution to improving AI systems. I will specify what I mean in future posts, but for now, I offer an introduction to how context is used in context-aware systems.
You end of April weekend read will teach you how to design better interfaces. Miller et. al., provides recommendations on three fronts: interface (presentation), Information (content), and interaction (function).
Which recommendation surprised you? As always, Tweet reply below to comment.
Sad news out of Tempe, Arizona: a self-driving care has struck and killed a pedestrian.
Self-driving fatalities will become more frequent in the years ahead. This is not a critique of the technology, but a hard truth. There are over 30,000 traffic related deaths in the United States every year. So, even if self-driving cars were 1,000 times safer than human drivers, they would still be involved in 30 deaths a year.
The current automotive death toll is important to keep in mind as we look at the safety record of self-driving cars. Would it be better to have self-driving cars kill 10,000 people a year or human-driven cars kill 30,000?
[Tweet-reply to comment]
At HIMMS, Jared Kuschner mentioned interoperability as a means to save lives. In addition, he posed customer (patient) ownership of healthcare data as a key principle of the Administration's strategy.
A new Weekend Reads is hot off the press: Quality and Efficiency of the Clinical Decision‐Making Process: Information Overload and Emphasis Framing. Laker et. al. look to address information overload through emphasis framing--highlighting some aspect of the information being exchanged. They report a laboratory study where a detail in the patient’s record was highlighted by bolding a line of text. Their results show that the highlighting significantly improved the quality of clinical evaluation. Interestingly, it also increased time to task completion. The authors speculated this was due to the highlighted detail leading to System 2 thinking because the correct plan of action for the simulated patient case is a divergence the usual plan of action.
The AWARE team is responsible for a number of different clinic decision support systems and dashboards that aid in patient care and are commercially available (Ambient Clinical Analytics). Last year they published two papers centered around electronic chart review in the ICU; one survey and one time motion study. This suggests the team is gearing up for their newest, cognitive load reducing product.
Are your ready for a challenge?
Challenges are a special type of post. The goal of a challenge is to crowdsource improvements to a proposed mechanism, Your input is critical because you might have the best opinion.
The below diagram shows that diet influences weight, influences your energy, influences your exercise, influences your disease risk. Exercise also feedback influences your diet.
What is missing? Is that arrow pointing the wrong direction?
Provide your input through a tweet reply below.