Loading
Nikhil Gopal
This is a live streamed presentation. You will automatically follow the presenter and see the slide they're currently on.
Nikhil Gopal
Definition of Bias
1. A systematic distortion of a statistical result due to a factor not allowed for in its derivation.
2. Prejudice in favor of or against one thing, person, or group compared with another, usually in a way considered to be unfair.
3. In some sports, such as lawn bowling, the irregular shape given to a ball.
Recognized Biases In Clinical Research
Interpretation Bias
Selection Bias
Publication Bias
Analysis Bias
Detection Bias
Exposure Bias
"I SPADE" (mnemonic device)
why recognize biases?
We cannot eliminate biases completely, but we can attempt to control for them
How would biases manifest in data visualizations?
"—the data, visual representation, textual annotations, and interactivity—and how visualizations denote and connote phenomena with reference to unstated viewing conventions and codes."
Visualization Rhetoric: Framing Effects in Narrative Visualization
Jessica Hullman & Nicholas Diakopoulos
IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS, VOL. 17, NO. 12, DECEMBER 2011
Data
Selecting which data to include and omit.
Examples: Excluding outliers.
Aggregating or summarizing data.
Reminescent of Analysis Bias
http://vudlab.com/simpsons/
http://vudlab.com/simpsons/
Visual Representation
How data dimensions are mapped to visual attributes.
Example: Continuous data loses resolution when mapped to grayscale
Reminiscent of Interpretation Bias (e.g. skewing retinal variables to support one perspective)
https://mycarta.files.wordpress.com/2011/11/render_compare_jet_detail.png
Annotation
Providing context via textual, graphical, or social means. Serves to focus reader attention.
Examples: Annotating a map with pertinent information. User comments under a blog post. Infographics!
Interactivity
Providing choices to a reader that constrain their exploration capabilities.
Example: An app with navigation views that allows a drill-down of only a certain class of data
But, wait a minute...some of these aren't exactly biases. They're design decisions!
But how do we differentiate?
Ask yourself:
Is this accurate?
Is this unfair?
Was this choice made for a good reason?
Also note: external constraints and limitations
The point is to provide a framework through which we can think about it!