Bias in algorithms

Conquering Data Bias

Konstantinos Kattidis

Data Analytics Lead

AI and data bias

AI has various use cases:

Personalized recommendations
Route navigation

What happens if algorithms become biased against certain groups, such as gender or race?

AI use cases

Algorithmic bias

Algorithmic bias arises when algorithms produce systematic and repeatable errors that result in unfair outcomes, favoring one group over another

It encompasses various types of biases that can emerge during the development and deployment of algorithms

Bias during algorithm training

Algorithmic bias is often initiated through bias in data collection
If the training data is not representative or is biased, the algorithm may learn and perpetuate those biases
For example: not having adequate representation of diverse groups in a facial recognition algorithm

Selection bias

Bias in feature selection

Feature selection bias

Feature selection bias happens when certain features are chosen for inclusion based on criteria that may lead to unfair or discriminatory outcomes
Such features are race, gender, or other protected characteristics
For example, an algorithm designed for predicting loan approvals may assume that income is the only relevant factor

Evaluation bias

Using a non-representative dataset during the testing phase of an algorithm can lead to evaluation bias
For example:
- A movie recommendation algorithm
- Testing performance based on only one genre
- Missing the fact that it might not work well for other genres like romance or comedy

Evaluation bias, testing only one genre group

Automation bias

Automation of MRI diagnosis

The tendency of individuals to rely too heavily on automated systems

For example:
- A medical diagnostic system is used to analyze medical images, such as MRIs
- Automation bias occur if the system's results are trusted without reviewing or confirming the findings
- This could lead to incorrect diagnoses

Let's practice!

Conquering Data Bias