Skip to content

ShilinSun/mxai_review

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 

Repository files navigation

Literature-MXAI-Review

This is the repository of A Review of Multimodal Explainable Artificial Intelligence: Past, Present and Future, a systematic review of Multimodal Explainable Artificial Intelligence (MXAI) studies in a historical perspective. For details, please refer to:

A Review of Multimodal Explainable Artificial Intelligence: Past, Present and Future Paper

🔥News

  • ☄️ [2024/12/26] We have revised the table of contents, and the remaining sections will be updated progressively.

Introduction

Artificial intelligence (AI) has rapidly developed through advancements in computational power and the growth of massive datasets. However, this progress has also heightened challenges in interpreting the "black-box" nature of AI models.

To address these concerns, eXplainable AI (XAI) has emerged with a focus on transparency and interpretability to enhance human understanding and trust in AI decision-making processes. In the context of multimodal data fusion and complex reasoning scenarios, the proposal of Multimodal eXplainable AI (MXAI) integrates multiple modalities for prediction and explanation tasks. Meanwhile, the advent of Large Language Models (LLMs) has led to remarkable breakthroughs in natural language processing, yet their complexity has further exacerbated the issue of MXAI.

To gain key insights into the development of MXAI methods and provide crucial guidance for building more transparent, fair, and trustworthy AI systems, we review the MXAI methods from a historical perspective and categorize them across four eras: traditional machine learning, deep learning, discriminative foundation models, and generative LLMs. We also review evaluation metrics and datasets used in MXAI research, concluding with a discussion of future challenges and directions.

We truly appreciate any contributions through PRs, issues, emails, or other means.

Contents

Related Surveys

Click to expand the XAI Survey Comparison Table
Ref. Published Year Literature coverage range Existing surveys are analyzed MXAI Transformer explainability LLMs explainability Historical perspective Data explainability Model explainability Post-hoc explainability Evaluation methods Main theme
Ours 2024 2000-2024 Historical perspective MXAI
Explainable Generative AI (GenXAI): A Survey, Conceptualization, and Research Agenda 2024 2016-2024 × × × × × × Explainable Generative AI
Usable XAI: 10 strategies towards exploiting explainability in the LLM era 2024 2017-2024 × × × × × LLM explainability strategies
Explainability for large language models: A survey 2024 2017-2024 × × × × × LLMs explainability
Rethinking Interpretability in the Era of Large Language Models 2024 2017-2024 × × × × × LLMs explainability
From Understanding to Utilization: A Survey on Explainability for Large Language Models 2024 2016-2023 × × × × × × LLMs explainability
Multimodal Explainable Artificial Intelligence: A Comprehensive Review of Methodological Advances and Future Research Directions 2023 2016-2023 × × × × × MXAI
Explainable Artificial Intelligence (XAI) from a user perspective: A synthesis of prior literature and problematizing avenues for future research 2023 2008-2022 × × × × × User and their concerns
A systematic review of Explainable Artificial Intelligence models and applications: Recent developments and future trends 2023 2018-2022 × × × × × × × XAI applications
Explainability of Vision Transformers: A Comprehensive Review and New Perspectives 2023 2016-2023 × × × × × Vision Transformer explainability
Explainability and Evaluation of Vision Transformers: An In-Depth Experimental Study 2023 2016-2023 × × × × × Vision Transformer explainability
Towards human-centered explainable ai: A survey of user studies for model explanations 2023 2018-2022 × × × × × Human-centered XAI
Explainable AI (XAI): A systematic meta-survey of current challenges and future opportunities 2023 2017-2021 × × × × × Challenges and trends in XAI
Explainable Artificial Intelligence (XAI): What we know and what is left to attain Trustworthy Artificial Intelligence 2022 2016-2022 × × × × Model’s trustworthiness
A survey on XAI and natural language explanations 2022 2006-2021 × × × × × × × Natural Language Explanations
Knowledge graphs as tools for explainable machine learning: A survey 2022 2015-2020 × × × × × × × × × Knowledge based XAI
Explainable AI methods-a brief overview 2022 2016-2021 × × × × × × × Introduction to XAI
Counterfactual explanations and how to find them: literature review and benchmarking 2022 2017-2022 × × × × × × Counterfactual explanations
Explainable AI for time series classification: a review, taxonomy and research directions 2022 2018-2022 × × × × × × XAI for time series
Unbox the black-box for the medical explainable AI via multi-modal and multi-centre data fusion: A mini-review, two showcases and beyond 2022 2018-2021 × × × × × × XAI in healthcare
A survey of contrastive and counterfactual explanation generation methods for explainable artificial intelligence 2021 1991-2020 × × × × × Contrastive and Counterfactual XAI
Notions of explainability and evaluation approaches for explainable artificial intelligence 2021 2015-2020 × × × × × × × Evaluation approaches of XAI
Explainable artificial intelligence: objectives, stakeholders, and future research opportunities 2021 2017-2020 × × × × × × × × × Black-box issue
Explainable ai: A review of machine learning interpretability methods 2021 2016-2020 × × × × × × × × ML interpretability methods
Explainable artificial intelligence approaches: A survey 2021 2016-2020 × × × × × × × XAI methods classification
Argumentation and explainable artificial intelligence: a survey 2021 2014-2020 × × × × × × × × Argumentation enabling XAI
A Survey on the Explainability of Supervised Machine Learning 2021 2015-2020 × × × × × × XAI methods classification
Reviewing the need for Explainable Artificial Intelligence (XAI) 2021 2016-2020 × × × × × × × × × Necessity of explainability
Explainable Artificial Intelligence (XAI): An Engineering Perspective 2021 2017-2020 × × × × × × × × × User and their concerns
What do we want from Explainable Artificial Intelligence (XAI)? – A stakeholder perspective on XAI and a conceptual model guiding interdisciplinary XAI research 2021 2016-2021 × × × × × × × × × User and their concerns
Interpretable deep learning: Interpretation, interpretability, trustworthiness, and beyond 2021 2016-2020 × × × × × × XAI methods classification

1. The Era of Traditional Machine Learning (2000-2009)

1.1 Data explainability

1.1.1 Feature selection methods

Filter
Wrapper
Embedded

1.1.2 Feature extraction methods

1.2 Model explainability

1.2.1 Linear/logistic regression

1.2.2 Decision Tree

1.2.3 K-Nearest Neighbors

1.2.4 Rule-based learning

1.2.5 Bayesian models

1.3 Post-hoc explainability

1.3.1 Model-agnostic techniques

Causal Explanation
Causal Prediction
Causal Intervention

1.3.2 Model-specific techniques

Tree ensemble
Support vector machines

2. The Era of Deep Learning (2011-2016)

2.1 Data explainability

2.1.1 Data quality analysis

2.1.2 Data interaction analysis

2.1.3 Inherently interpretable models

2.2 Model explainability

2.2.1 Deep neural network interpretability

Decomposability
Algorithmic transparency

2.2.2 Explain the training process

Attention-based networks
Disentangled representations
Generate explanations

2.3 Post-hoc explainability

2.3.1 Multi-layer neural networks

Model simplification
Feature-related explanations

2.3.2 Convolutional Neural Networks

Understand decision-making processes
Investigate module function

2.3.3 Recurrent Neural Networks

Feature-related explanations
Local explanations

3. The Era of Discriminative Foundation Models (2017-2021)

3.1 Data explainability

3.1.1 Analyse multimodal datasets

3.1.2 Structural relationship construction

3.2 Model explainability

3.2.1 Behavioral explanation

Architecture-independent
Transformer-based
CLIP-based

3.2.2 Structural transparency

GNN-based
Knowledge Graph-based
Causal-based
Others

3.3 Post-hoc explainability

3.3.1 Counterfactual-based

3.3.2 Bias mitigation

3.3.3 Multimodal learning process explanation

Multimodal representation
Multimodal reasoning
Visualization

4. The Era of Generative Large Language Models (2022-2024)

4.1 Data explainability

4.1.1 Explain datasets

4.1.2 Data selection

4.1.3 Graph modeling

4.2 Model explainability

4.2.1 Process explanation

Process explanation
Explain multimodal-ICL
Explain multimodal-CoT
Robustness enhancement

4.2.2 Explainable data augmentation

Explainable data augmentation
Small models training

4.2.3 Inherent interpretability

Inherent interpretability
Image-Text understanding
Video-Text understanding
Audio-Text understanding
Multimodal-Text understanding
Classifier-based

4.3 Post-hoc explainability

4.3.1 Probing-based explanation

Probing-based explanation
Parameter-free
Modular design-based

4.3.2 Reasoning-based explanation

Reasoning-based explanation
External world knowledge-based
Feedback-based

4.3.3 Example-based explanation

Example-based explanation
Counterfactual examples
Adversarial examples

5. Evaluation datasets and metrics

Citation

If you find this repository useful, please cite our paper:

@article{sun2024review,
  title={A Review of Multimodal Explainable Artificial Intelligence: Past, Present and Future},
  author={Sun, Shilin and An, Wenbin and Tian, Feng and Nan, Fang and Liu, Qidong and Liu, Jun and Shah, Nazaraf and Chen, Ping},
  journal={arXiv preprint arXiv:2412.14056},
  year={2024}
}

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published