Unraveling the data dilemma

Conquering Data Bias

Konstantinos Kattidis

Data Analytics Lead

Data driving decision-making

Data-driven decision-making has revolutionized industries across the globe
The adoption of data and AI has extended its reach into critical areas like:
- Social justice
- Employment screening
- Interactive interfaces such as virtual assistants

Use cases of data driven decision making

The emergence of data bias

Alongside these advancements, concerns have emerged regarding the presence of biases within these solutions.

CV Screening

Amazon shut down its model that scored suitable candidates for employment as it favored male candidates

What is data bias?

It arises when data or information is limited, painting an inaccurate or unfair picture of the population
It results from various factors such as:
- Imbalances and underrepresentation
- Historical prejudices
- Cognitive tendencies
It compromises the accuracy and reliability of data, hindering its effectiveness in informing decision-making processes

The origins of data bias

The Amazon case

Amazon's model was trained on resumes submitted over 10-years.
During this time, male candidates were disproportionately hired or favored by the hiring process.
The model was biased towards male candidates due to historical hiring patterns.
Addressing data bias requires scrutiny of both the dataset quality and the human processes involved.

Amazon boxes and cv screening

About the course

Conquer the bias

Chapter 1 - Learn about the impact of data bias on decision-making and why it is important to take data bias into consideration when working with data.

Chapter 2 - Learn the various types of data bias in data collection and discover how to implement techniques to minimize it.

Chapter 3 - Learn the data bias types commonly present during data analysis and when sharing analysis results. We will dive into how to identify bias and how to mitigate it.

Let's practice!

Conquering Data Bias