Awesome Human-Centered AI (HCAI)

Research Papers

• Data collection and annotation
• Weak supervision and self supervision
• Explainable ML
• Debate on explainable ML
• AI bias (Dataset bias and Algorithmic bias)
• AI robustness
• Human-AI Collaboration
• Human-AI Creation
• Human-in-the-loop autonomy
• Superhuman AI and knowledge generation

Data collection and annotation

• Crowdsourcing in Computer Vision, Foundations and Trends in Computer Graphics and Vision, 2016
• Efficient Interactive Annotation of Segmentation Datasets with Polygon-RNN++, CVPR, 2018 | code
• RoboTurk: A Crowdsourcing Platform for Robotic Skill Learning through Imitation, CoRL, 2018 | code
• From ImageNet to Image Classification: Contextualizing Progress on Benchmarks, ICML, 2020 | code

Weak supervision and self supervision

• What's the Point: Semantic Segmentation with Point Supervision, ECCV, 2016 | code
• Revisiting Unreasonable Effectiveness of Data in Deep Learning Era, ICCV, 2017
• Weakly Supervised Object Localization Papers
• TS-CAM: Token Semantic Coupled Attention Map for Weakly Supervised Object Localization, ICCV, 2021 | code
• DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision, ICCV, 2021 | code
• Universal Weakly Supervised Segmentation by Pixel-to-Segment Contrastive Learning, ICLR, 2021 | code
• Language-driven Semantic Segmentation, ICLR, 2022 | code
• Masked Autoencoders Are Scalable Vision Learners, CVPR, 2022 | code

Explainable ML

• Understanding the role of individual units in a deep neural network, PNAS, 2020 | code
• Feature Visualization, Distill, 2017
• The Building Blocks of Interpretability, Distill, 2018
• CAM, and many CAM variants (Grad-CAM, Grad-CAM++, Score-CAM, Ablation-CAM and XGrad-CAM)
• Towards Automatic Concept-based Explanations, NeurIPS, 2019 | code

Debate on explainable ML

• The Mythos of Model Interpretability, ICML, 2016
• Sanity Checks for Saliency Maps, NeurIPS, 2018 | code
• On the importance of single directions for generalization, ICLR, 2018 | code
• Revisiting the Importance of Individual Units in CNNs via Ablation, arXiv, 2018
• Towards falsifiable interpretability research, NeurIPS, 2020
• The false hope of current approaches to explainable artificial intelligence in health care, The Lancet, 2021
• Post hoc Explanations may be Ineffective for Detecting Unknown Spurious Correlation, ICLR, 2022

AI bias (Dataset bias and Algorithmic bias)

• Unbiased Look at Dataset Bias, CVPR, 2011
• Women also Snowboard: Overcoming Bias in Captioning Models, ECCV, 2018
• Moving beyond “algorithmic bias is a data problem”, Patterns, 2021
• Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification
• Algorithmic bias detection and mitigation: Best practices and policies to reduce consumer harms

AI robustness

• Physical Adversarial Examples for Object Detectors, USENIX WOOT 2018
• Towards Robust LiDAR-based Perception in Autonomous Driving: General Black-box Adversarial Sensor Attack and Countermeasures, USENIX Security 2020
• Adversarial Examples Are Not Bugs, They Are Features, NeurIPS, 2019 | code
• Benchmarking Neural Network Robustness to Common Corruptions and Perturbations, ICLR, 2019 | code
• Noise or Signal: The Role of Image Backgrounds in Object Recognition, ICLR, 2021 | code
• ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness, ICLR, 2019 | code
• Learning Independent Causal Mechanisms, ICML, 2018 | code

Human-AI Collaboration

• Human-Centered Tools for Coping with Imperfect Algorithms during Medical Decision-Making, CHI, 2019
• Human–computer collaboration for skin cancer recognition, Nature Medicine, 2020
• To Trust or to Think: Cognitive Forcing Functions Can Reduce Overreliance on AI in AI-assisted Decision-making, ACM HCI, 2021
• OpenAI Codex

Human-AI Creation

• Semantic Image Synthesis with Spatially-Adaptive Normalization, CVPR, 2019 | code
• GANSpace: Discovering Interpretable GAN Controls, NeurIPS, 2020 | code
• Sketch Your Own GAN, ICCV, 2021 | code
• LatentCLR: A Contrastive Learning Approach for Unsupervised Discovery of Interpretable Directions, ICCV, 2021 | code
• LayoutGAN: Synthesizing Graphic Layouts With Vector-Wireframe Adversarial Networks, TPAMI, 2021
• Closed-Form Factorization of Latent Semantics in GANs, CVPR, 2021 | code
• Make-A-Scene: Scene-Based Text-to-Image Generation with Human Priors, arXiv, 2022 | code

Human-in-the-loop autonomy

• Deep reinforcement learning from human preferences, NeurIPS, 2017 | code
• Shared Autonomy via Deep Reinforcement Learning, RSS, 2018 | code
• Understanding RL Vision, Distill, 2020
• Learning from Interventions: Human-robot interaction as both explicit and implicit feedback, RSS, 2020
• PEBBLE: Feedback-Efficient Interactive Reinforcement Learning via Relabeling Experience and Unsupervised Pre-training, ICML, 2021 | code
• Pragmatic Image Compression for Human-in-the-Loop Decision-Making, NeurIPS, 2021 | code
• Recent advances in leveraging human guidance for sequential decision-making tasks
• Efficient Learning of Safe Driving Policy via Human-AI Copilot Optimization , ICLR, 2022

Superhuman AI and knowledge generation

• Grandmaster level in StarCraft II using multi-agent reinforcement learning, Nature, 2019
• Acquisition of Chess Knowledge in AlphaZero, arXiv, 2021
• Outracing champion Gran Turismo drivers with deep reinforcement learning, Nature, 2022

HCAI Courses

• HCAI for CV
  • Human-Centered AI for Computer Vision and Machine Autonomy, UCLA
• HCAI for NLP
  • Human-Centered Machine Learning, UChicago
• HCAI for Human-Computer Interaction (HCI)
  • Human-AI Interaction, CMU
  • Human-Computer Interaction, UCLA

HCAI Workshop

• HCAI, NeurIPS2021
• Human-Centered Explainable AI (HCXAI), CHI 2022
• Human-Centered AI for Computer Vision, CVPR 2022

Acknowledgement

In this repo, currrent module list and paper list mainly based on Bolei Zhou's HCAI course at UCLA, feel free to add any new papers or modules!