General Biases¶
Historical Bias¶
Definition
Historical biases exist prior to the inception of any data science project. However, they arise within the context of a project when there is a gap or misalignment between the state of the world and the goals or objectives that are encoded and propagated in the model or system being developed. For instance, even with perfect sampling and feature selection, a model that perpetuates (or exacerbates) socioeconomic inequalities through the outcomes it promotes is contributing to historical bias.
Deliberative Prompts
Which groups and communities will be affected by the use of your model?
Are there groups or communities that will be excluded from your model? If so, why?
Is there a risk of worsening or perpetuating socioeconomic inequalities in the development and deployment of your model?
Confirmation Bias¶
Definition
A tendency to search for, gather, or use information that confirms pre-existing ideas and beliefs, and to dismiss or downplay the significance of information that disconfirms one’s favoured hypothesis. This can be the result of motivated reasoning or sub-conscious attitudes, which in turn may lead to prejudicial judgements that are not based on reasoned evidence.
Deliberative Prompts
What mechanisms do you have in place within your team that can help ensure a diversity of viewpoints that may mitigate the effects of confirmation bias?