GPH FINAL

Bias device -Kadambi

one that operates in a manner that disadvantages certain demographic

groups and influences health inequity (three primary forms)

three types of medical device bias -Kadambi

Physical bias: Physical working principle of a medical device is bias when it exhibits an undesirable performance across demographic groups

Computational bias: bias that invades medical technology

Interpretation bias: when a medical device is subject to bias inference of readings (spirometer)

Dataset bias -Kadambi

An aspect of computational bias, algorithms can learn from imbalanced datasets where certain groups of people are underrepresented in the sample

5 types of algorithmic bias -Muralidharan

representation bias, omitted variable bias, aggregation bias, linking bias, label bais

Representation bias -Muralidharan

Data underrepresents or misrepresents subsets of the population measurement bias—data inaccurately reflects the variables

Omitted variable bias -Muralidharan

Omitted data. For example, the“Black-box” nature of many algorithms makes it difficult to figure out which variables were used and omitted.

Aggregation Bias -Muralidharan

conclusions are drawn about individuals based on observations about a larger group

Linking Bias: -Muralidharan

correlations that AI draws about particular users based on the characteristics of other users that may not be accurate in a heterogeneous population

Label Bias -Muralidharan

Bias that arises when the same object is labeled differently based on differing perceptions by the annotator

Stratification approach

medical devices designed and tailored to fit different groups

- Categorizing people into groups poses new (moral) challenges, reinforce stereotypes

One-size-fits-all approach -de Kanter

allows for a single off-the-shelf available product; assumes a certain end-user and caters to that typical user in the design and testing process (normalization)

Morally significant difference -de Kanter

distinction between two actions, cases, or policies that alters their moral permissibility or weight

Pulse oximeter

uses optical biosensors that use light to monitor vital signs, this is known to not work well on darker skin tones and can lead to delay in treatment

Algorithm bias

when the mathematics of data processing disadvantages certain groups