Eerlijkheid in AI-projecten

Je maakt deel uit van een team dat een AI-toelatingssysteem voor universiteiten ontwikkelt. Dit systeem wil aanmelders eerlijk beoordelen op basis van verschillende factoren. Het team overweegt drie verschillende benaderingen om eerlijkheid in het project te beoordelen, waaronder disparate impact, gelijke kansen en demografische pariteit.

Deze oefening maakt deel uit van de cursus

Verantwoord AI-gegevensbeheer

Cursus bekijken

Praktische interactieve oefening

Zet theorie om in actie met een van onze interactieve oefeningen.

Begin met trainen

Deze oefening maakt deel uit van de cursus

Verantwoord AI-gegevensbeheer

SkillTag.level.intermediateSkillTag.label

4.7+

Begin de cursus gratis

Learn about the fundamental theory behind responsible data management in AI. You’ll review key dimensions such as security, transparency, fairness, and more before conceptualizing the metrics and challenges associated with these dimensions and understanding how to balance responsible AI with other business and technical requirements.

Exercise 1: Dimensies van verantwoord omgaan met data Exercise 2: AI verantwoord ontwikkelen Exercise 3: Dimensies van verantwoord omgaan met data Exercise 4: KPI's voor verantwoorde AI Exercise 5: Een verantwoordelijk AI-project plannen Exercise 6: Verantwoord datagebruik Exercise 7: Eerlijkheid in AI-projecten

Huidige oefening

Exercise 8: Uitdagingen van verantwoorde AI Exercise 9: Afwegingen in verantwoorde AI Exercise 10: Professionele plichten en ethisch handelen

Data regulation is essential to the legality of any AI project. Learn about key regulations, third-party licenses, and compliance strategies for informed consent and data-sharing agreements (with legal counsel). Finally, you'll learn about developing robust data governance strategies and management plans to ensure your project remains compliant throughout its lifecycle.

Exercise 1: Overview of data regulation Exercise 2: Data protection laws Exercise 3: Data regulation in a project Exercise 4: Data compliance Exercise 5: Data owner rights and compliance Exercise 6: Data use agreements (DUAs)Exercise 7: Third-party licensing Exercise 8: Types of licenses Exercise 9: Licensing agreements Exercise 10: Selecting a license Exercise 11: Data governance and data management plan (DMP)Exercise 12: Data management Exercise 13: Components of a DMP

Navigate through the responsible selection and integration of data sources by understanding the importance of data origin, nature, and temporality, emphasizing legal compliance, diversity, and fairness. By exploring types of bias and their origins, you’ll look at data fairness and representation to create a comprehensive dataset for modeling.

Exercise 1: Data sources Exercise 2: Data source types Exercise 3: Data source and responsible dimensions Exercise 4: Data source limitations Exercise 5: Limitations in data sources Exercise 6: Data sources and bias Exercise 7: Data source selection Exercise 8: Evaluation of data sources Exercise 9: Data augmentation Exercise 10: Data integration Exercise 11: Data integration steps Exercise 12: Risks and benefits of data integration

Understand data audits, data validation, and bias mitigation. Data pre-processing and catching bias in modeling do not sound like fun, but let's streamline them with common approaches and trusted techniques!

Exercise 1: Data audit Exercise 2: Data audit in the project lifecycle Exercise 3: Need for data audit Exercise 4: Data validation Exercise 5: Defining data validation Exercise 6: Data validation approaches Exercise 7: Data validation best practices Exercise 8: Subgroup analysis Exercise 9: Data validation in the project Exercise 10: Pre-processing and bias Exercise 11: Bias mitigation Exercise 12: Mitigation strategies Exercise 13: Bias mitigation throughout project lifecycle Exercise 14: Consequences of bias mitigation Exercise 15: Post-deployment bias Exercise 16: Congratulations!