mlsanityData to 7 Oct 2026

Exploring the Lands Between: A Method for Finding Differences between AI-Decisions and Human Ratings through Generated Samples

arXiv 2409.12801 ↗First posted 19 Sep 2024cs.HC cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Many important decisions in our everyday lives, such as authentication via biometric models, are made by Artificial Intelligence (AI) systems. These can be in poor alignment with human expectations, and testing them on clear-cut existing data may not be enough to uncover those cases. We propose a method to find samples in the latent space of a generative model, designed to be challenging for a decision-making model with regard to matching human expectations. By presenting those samples to both the decision-making model and human raters, we can identify areas where its decisions align with human intuition and where they contradict it. We apply this method to a face recognition model and collect a dataset of 11,200 human ratings from 100 participants. We discuss findings from our dataset and how our approach can be used to explore the performance of AI models in different contexts and for different user groups.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 30 of its 42 references that are in the map. First posted on arXiv 19 Sep 2024.

Builds on

30 of 42 references.

References · 30

  1. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  2. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  3. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  4. 2018A Style-Based Generator Architecture for Generative Adversarial Networkscited by 5,986
  5. 2021Zero-Shot Text-to-Image Generationcited by 4,276
  6. 2019Analyzing and Improving the Image Quality of StyleGANcited by 3,580
  7. 2015FaceNet: A Unified Embedding for Face Recognition and Clusteringcited by 2,833
  8. 2015Variational Inference with Normalizing Flowscited by 2,161
  9. 2019Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AIcited by 1,543
  10. 2018Glow: Generative Flow with Invertible 1x1 Convolutionscited by 1,539
  11. 2019Natural Adversarial Examplescited by 1,401
  12. 2017StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translationcited by 1,036
18 more
  1. 2018A Survey Of Methods For Explaining Black Box Modelscited by 970
  2. 2016The Mythos of Model Interpretabilitycited by 716
  3. 2020FastSpeech 2: Fast and High-Quality End-to-End Text to Speechcited by 710
  4. 2019Interpreting the Latent Space of GANs for Semantic Face Editingcited by 691
  5. 2020NVAE: A Deep Hierarchical Variational Autoencodercited by 635
  6. 2020GANSpace: Discovering Interpretable GAN Controlscited by 612
  7. 2023T2M-GPT: Generating Human Motion from Textual Descriptions with Discrete Representationscited by 577
  8. 2020Evaluation: from precision, recall and F-measure to ROC, informedness, markedness and correlationcited by 420
  9. 2019On the "steerability" of generative adversarial networkscited by 247
  10. 2020Adversarial Latent Autoencoderscited by 155
  11. 2018Visual Analytics in Deep Learning: An Interrogative Survey for the Next Frontierscited by 117
  12. 2019GANSynth: Adversarial Neural Audio Synthesiscited by 114
  13. 2019FairVis: Visual Analytics for Discovering Intersectional Bias in Machine Learningcited by 93
  14. 2021A Comprehensive Study on Face Recognition Biases Beyond Demographicscited by 86
  15. 2020Towards causal benchmarking of bias in face analysis algorithmscited by 64
  16. 2021On the use of automatically generated synthetic image datasets for benchmarking face recognitioncited by 26
  17. 2019Explaining Vulnerabilities to Adversarial Machine Learning through Visual Analyticscited by 23
  18. 2021Attribute-specific Control Units in StyleGAN for Fine-grained Image Manipulationcited by 13

Built on by

Not cited yet.