According to the chapter, institutions using AI should overcome algorithmic bias by choosing the correct learning models, using the right training data set, performing meaningful data processing, monitoring real-world performance across the AI lifecycle, and avoiding infrastructural issues.

The chapter explicitly lists approaches from Shimasaan and Bukohwo (2015): selecting correct learning models, using the right training data set, performing data processing meaningfully, monitoring real-world performance across the AI lifecycle, and avoiding infrastructural issues. It also recommends applying appropriate methods to process, analyze, and interpret data, continuously monitoring and updating algorithms to prevent bias and ensure fairness, developing more robust fairness metrics, and researching ways to improve transparency and explainability in AI decision-making. Institutions are further urged to review data collection methods, adopt ethical codes, employ experts to monitor ethical AI use, and establish ethics committees to oversee AI in democratic processes.

Key points

Choose correct or appropriate learning models.
Use the right training data set.
Perform meaningful data processing and use appropriate analysis methods.
Monitor real-world performance across the AI lifecycle and update algorithms continuously.
Avoid infrastructural issues that can affect AI performance.
Develop robust fairness metrics and improve transparency and explainability.
Adopt ethical data collection codes, employ expert oversight, and establish ethics committees.

According to the chapter, what are the specific approaches that institutions using AI should apply to overcome algorithmic bias?

Related questions