• / social bias
  • / algorithmic bias
  • / labeling
  • / machine learning
  • / AI
  • / human
  • / social bias
  • / algorithmic bias
  • / labeling
  • / machine learning
  • / AI
  • / human /
    • "The unplanned
  • bias
    • of
  • machine learning
    • "
  • / algorithmic bias
  • machine learning
  • / AI
  • / human
  • / social bias
  • / algorithmic bias
  • / labeling
  • / social bias
  • / algorithmic bias
  • / labeling
  • / AI
  • / human /
  • #SUMMARY:
    • “The real hazard in
  • machine learning
    • has less to do with robotic conscious entities and more to do with another type of conscious entity -
  • human beings.
    • When societally
  • biased data
    • is used to train a
  • learning
    • model, the insidious outcome is a discriminatory
  • machine learning
    • model that predicts the
  • societal biases
    • we aim to eliminate”.[1]
  • / social bias
  • / algorithmic bias
  • / labeling
  • / machine learning
  • / AI
  • / human /
    • “
  • Humans
    • are error-prone and
  • biased
    • , but that doesn’t mean that
  • algorithms
    • are necessarily better.”[2]
  • / social bias
  • / algorithmic bias
  • / labeling
  • / machine learning
  • / AI
  • / human /
    • “Training sets, then, are the foundation on which contemporary
  • learning
    • systems are built. They are central to how
  • AI
    • systems recognize and interpret the world” [3]
  • / machine learning
  • / AI
  • / human /
  • / social bias
  • / algorithmic bias
  • / labeling
  • / social bias
  • / algorithmic bias
  • / labeling
  • / machine learning
  • / AI
  • / human /
  • #RESEARCH
  • DIRECTION ->
    • Purpose: > Promote an informed reflection on
  • human labeling
    • and how this
  • bias
    • affects and influences the
  • datasets
    • that are fed by humans into the various
  • machine learning
  • algorithms
    • . - ("biased data in, biased data out"); / > Present, raise awareness and print transparency about some examples of "
  • human bias
    • " (ImageNet and worldNet) and understand what are the implications of these in the propagation of
  • social
    • injustices (Stereotypes, prejudices, gender/ethnic discrimination, etc...).

  • / machine learning
  • / AI
  • / human
  • / social bias
  • / algorithmic bias
  • / labeling
  • / social bias
  • / algorithmic bias
  • / labeling
  • / machine learning
  • / AI
  • / human /
    • Objectives: >" Denounce" the parallelism of the
  • biases
    • that have been carried over from
  • human
    • to
  • machine-learning
    • . / > In the web component, systematically build a dynamic and effusive
  • database
    • , in order to elucidate and expose/report the problem of
  • labeling
  • bias
    • in
  • machine learning
  • algorithms
    • ; / > As a complement to the web (print), to create an educational lexicon that explains the meaning of the most relevant
  • biases
    • , i.e. those that directly affect decision making.
  • / machine learning
  • / AI
  • / human
  • / social bias
  • / algorithmic bias
  • / labeling
  • / social bias
  • / algorithmic bias
  • / labeling
  • / machine learning
  • / AI
  • / human /
    • Metodology: > Content gathering: selection and definition of content for the web (
  • bias
    • related / projects / examples + quotes) and for the editorial (
  • bias
    • lexicon); / > Development of the web platform, compile and select effects that convey the theme in an expressive, experimental and dynamic way (, js, p5js,etc.); / > Development of the print component in a coherent, explanatory and affirmative way;
  • / machine learning
  • / AI
  • / human
  • / social bias
  • / algorithmic bias
  • / labeling
  • / social bias
  • / algorithmic bias
  • / labeling
  • / machine learning
  • / AI
  • / human /
  • #RESULTS PREVIEW:
    • > Web component: allow reflection and raise awareness of the negative impacts that
    • data
  • labeling bias
    • have on
  • machine learning
  • learning
  • algorithms
    • and consequently on society; / > Editorial complement: will refer to the various types of
  • bias
    • : aims to explain and/or exemplify the most influential and significant types of
  • bias
    • .
  • / machine learning
  • / AI
  • / human /
  • / social bias
  • / algorithmic bias
  • / labeling
  • / social bias
  • / algorithmic bias
  • / labeling
  • / machine learning
  • / AI
  • / human /
    • “
  • machine learning
    • capitalizes on
    • (or “learn” from) statistical regularities in
    • -generated
  • datasets
    • , tends to pick up
  • social
    • patterns that manifest in
  • human
    • behavior and that are reflected in the data on which it is trained.”[4]

[1] https://towardsdatascience.com/our-machine-learning-algorithms-are-magnifying-bias-and-perpetuating-social-disparities-6beb6a03c939
[2] https://www.vox.com/recode/2020/2/18/21121286/algorithms-bias-discrimination-facial-recognition-transparency
[3] https://excavating.ai/
[4] http://philsci-archive.pitt.edu/17169/1/Algorithmic%20Bias.pdf








Marco Alpoim | MDC 20/21 | Projeto II | Laboratório II