Trusted AI
Our trust in technology relies on understanding how it works. We need to understand why AI makes the decisions it does. We're developing tools to make AI more explainable, fair, robust, private, and transparent.
Overview
When people understand how technology works, and we can asses that it’s safe and reliable, we’re far more inclined to trust it. Many AI systems to date have been black boxes, where data is fed in and results come out. To trust a decision made by an algorithm, we need to know that it is fair, that it’s reliable and can be accounted for, and that it will cause no harm. We need assurances that AI cannot be tampered with and that the system itself is secure. We need to be able to look inside AI systems, to understand the rationale behind the algorithmic outcome, and even ask it questions as to how it came to its decision.
At IBM Research, we’re working on a range of approaches to ensure that AI systems built in the future are fair, robust, explainable, account, and align with the values of the society they’re designed for. We’re ensuring that in the future, AI applications are as fair as they are efficient across their entire lifecycle.
Teams
- AI TestingWe’re designing tools to help ensure that AI systems are trustworthy, reliable and can optimize business processes.
- Adversarial Robustness and PrivacyWe’re making tools to protect AI and certify its robustness, and helping AI systems adhere to privacy requirements.
- Explainable AIWe’re creating tools to help AI systems explain why they made the decisions they did.
- Fairness, Accountability, TransparencyWe’re developing technologies to increase the end-to-end transparency and fairness of AI systems.
- Trustworthy GenerationWe’re developing theoretical and algorithmic frameworks for generative AI to accelerate future scientific discoveries.
- Uncertainty QuantificationWe’re developing ways for AI to communicate when it's unsure of a decision across the AI application development lifecycle.
Tools + code
ART: Adversarial Robustness Toolbox
A Python library for machine learning security that enables developers and researchers to defend and evaluate machine learning models and applications against the adversarial threats of evasion, poisoning, extraction, and inference.
View project →AI Privacy 360
Tools to support the assessment of privacy risks of AI-based solutions, and to help them adhere to any relevant privacy requirements. Tradeoffs between privacy, accuracy, and performance can be explored at different stages in the machine learning lifecycle.
View project →AI Explainability 360
This open source toolkit contains eight algorithms that help you comprehend how machine-learning models predict labels throughout the AI application lifecycle. It’s designed to translate algorithmic research into the real-world use cases in a range of files, such as finance, human capital management, healthcare, and education.
View project →AI Fairness 360
An open-source toolkit of metrics to check for unwanted bias in datasets and machine learning models, and state-of-the-art algorithms to mitigate such bias. Containing over 70 fairness metrics and 10 bias mitigation algorithms, it’s designed to turn fairness research into practical applications.
View project →AI FactSheets 360
Toolkit to create factsheets outlining the details about how an AI service operates, how it was trained and tested, its performance metrics, fairness and robustness checks, intended uses, maintenance, and other critical details.
View project →Uncertainty Quantification 360
An open-source Python package that provides a diverse set of algorithms to quantify uncertainty, as well as capabilities to measure and improve UQ to streamline the development process.
View project →Causal Inference 360
A Python package for modular causal inference analysis and model evaluations.
View project →
Science for Social Good
IBM Science for Social Good partners IBM Research scientists and engineers with academic fellows, subject matter experts from NGOs, public sector agencies, and social enterprises to tackle emerging societal challenges using science and technology.
Publications
- 2022
- IJCAI 2022
- 2022
- CVPR 2022
- 2022
- ACL 2022
- 2022
- ACL 2022
- 2022
- ICDE 2022
- 2022
- arXiv
Building trustworthy AI with Watson
Our research is regularly integrated into Watson solutions to make IBM’s AI for business more transparent, explainable, robust, private, and fair.
Our work
What is human-centered AI?
ExplainerNew AI research could help find the best treatment for bowel disease
ResearchAccelerating molecular optimization with AI
Deep DiveSecuring AI systems with adversarial robustness
Deep DiveProject Debater argues for key point analysis of surveys, reviews, and social media
ReleaseEvaluating common sense in AI
Deep Dive- See more of our work on Trusted AI
