Accountability, Evaluation, and Obscurity of AI Algorithms

Permanent URI for this collection

https://hdl.handle.net/10125/102478

Browse

Recent Submissions

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Applying Reflexivity to Artificial Intelligence for Researching Marginalized Communities and Real-World Problems
    (2023-01-03) Aguilar, Nathan; Landau, Aviv; Mathiyazhagan, Siva; Auyeung, Alex; Dillard, Sarah; Patton, Desmond
    Despite advances in artificial intelligence (AI), ethical principles have been overlooked, harming marginalized communities. These flaws are due to a lack of critical insight into the complex positionality of the researcher, power dynamics between scholars and the communities being studied, and the structural impact on real-world problems when AI systems appear to be accurate but ethically fail. Reflexivity is a process that yields a better understanding of community-specific nuances, areas requiring local expertise, and the potential consequences of scholastic interventions for real-world problems (i.e., social, environmental, or socioeconomic). The paper builds on the five stages of social work reflexivity that can be applied to AI researchers and provided questions that can be asked in order to increase privacy, accountability, and fairness. We discuss the effective implementation of reflexivity in research, detail the stages of social work reflexivity and highlight key questions for AI researchers to ask throughout the research process.
  • Thumbnail Image
    Item
  • Thumbnail Image
    Item
    Comparing Methods for Mitigating Gender Bias in Word Embedding
    (2023-01-03) Ronchieri, Elisabetta; Biagi, Clara
    Word embedding captures the semantic and syntactic meaning of words into dense vectors. It contains biases learning from data that include constructs, cultural stereotypes, and inequalities of the society. Many methods for removing bias in traditional word embedding have been proposed. In this study we use the original GloVe word embedding and perform a comparison among debiasing methods built on top of GloVe in order to determine which methods perform the best removing bias. We have defined half-sibling regression, repulsion attraction neutralization GloVe method and compared it with gender-preserving, gender-neutral GloVe method and other debiased methods. According to our results, no methods outperform in all the analyses and completely remove gender information from gender neutral words. Furthermore, all the debiasing methods perform better than the original GloVe.