What is Responsible AI?
Nokia has defined six principles, or pillars, that should guide all AI research and development in the future. We believe these principles should be applied the moment any new AI solution is conceived and then enforced throughout its development, implementation and operation stages.
Nokia Bell Labs and e& announce R&D collaboration to innovate for strategic industrial sectors
The goal is to develop responsible AI solutions for sustainable enterprise and industrial automation applications and accelerate innovation concepts toward real world deployments.
For AI to thrive, it must put humans first
Responsible AI research and vision
Video
Testing AI’s limits: Mimicking Human and Analytical Thinking
Blog
Blog
Blog
Video
Why we need Responsible AI
White paper
Six pillars of Responsible AI
Fairness
AI systems must be designed in ways that maximize fairness, non-discrimination and accessibility. All AI designs should promote inclusivity by correcting both unwanted data biases and unwanted algorithmic biases.
Designing Value-Aligned Human-AI Interaction
Zana Buçinca examines the over- or under-reliance on AI advice due to design flaws, advocating for AI tools that align with human values to improve decision-making and human…
Responsible AI in Education
Rene Kizilcec explores AI in education, focusing on trust, cultural bias, and improving AI models as personalized tutors, with promising examples and future research…
Emotion AI in the Future of Work
Nazanin Andalibi critiques Emotion AI's rise in workplaces, discussing validity, bias, and surveillance concerns. She highlights its sociotechnical impact and persistent harms…
Investigating Algorithmic Biases in Child Welfare Systems Through Human-Centered Data Science
Shion Guha's study reveals biases in child-welfare algorithms, highlighting flawed data use and legal-risk focus over maximizing children's welfare outcomes.
The Ethics of Emotion in Artificial Intelligence Systems
Artificial Emotional Intelligence raises ethical concerns. Luke questions emotion modeling and the appropriateness of deploying AEI for public use.
Human-Centered Approaches to Supporting Fairness in AI
Vivek Krishnamurthy of University of Ottawa gave a talk titled “Human-Centered Approaches to Supporting Fairness in AI”.
The Future of AI for Social Good
Saiph Savage demonstrates how AI-based coaching on crowd-sourcing platforms can boost worker wages, efficiency, and skills while promoting fairness.
Ethics in AI: A Challenging Task
Ricardo Baeza-Yates from Institute for Experiential AI at Northeastern University gave a talk titled “Ethics in AI: A Challenging Task”.
Human-Centered Approaches to Supporting Fairness in AI
Michael Madaio from Microsoft Research discusses human-centered approaches to fairness in AI, focusing on context-driven checklists across AI design phases.
Designing Artificial Intelligence to Navigate Societal Disagreement
Michael Bernstein of Stanford introduced 'jury learning,' a method addressing bias in AI by ensuring underrepresented groups influence classifier predictions.
Maintaining fairness under distribution shift
Jessica Schrouff from Google Research explores the problem of distribution shift in healthcare AI, focusing on dermatology and electronic health records.
Is Legal AI Ethical AI?
Harvard researchers discuss the ethical challenges of legal AI, exploring how moral principles are translated into algorithms, and the risks of unintended discrimination.
Who are we listening to? Building blocks for trustworthy AI
Hillary Juma of Mozilla Foundation introduces the Common Voice project, improving voice tech accessibility by collecting 13,000 hours of data in 76 languages.
Reliability, Safety and Security
Privacy
Transparency
Sustainability
Accountability
AI systems must be designed in ways that maximize fairness, non-discrimination and accessibility. All AI designs should promote inclusivity by correcting both unwanted data biases and unwanted algorithmic biases.
Designing Value-Aligned Human-AI Interaction
Zana Buçinca examines the over- or under-reliance on AI advice due to design flaws, advocating for AI tools that align with human values to improve decision-making and human…
Responsible AI in Education
Rene Kizilcec explores AI in education, focusing on trust, cultural bias, and improving AI models as personalized tutors, with promising examples and future research…
Emotion AI in the Future of Work
Nazanin Andalibi critiques Emotion AI's rise in workplaces, discussing validity, bias, and surveillance concerns. She highlights its sociotechnical impact and persistent harms…
Investigating Algorithmic Biases in Child Welfare Systems Through Human-Centered Data Science
Shion Guha's study reveals biases in child-welfare algorithms, highlighting flawed data use and legal-risk focus over maximizing children's welfare outcomes.
The Ethics of Emotion in Artificial Intelligence Systems
Artificial Emotional Intelligence raises ethical concerns. Luke questions emotion modeling and the appropriateness of deploying AEI for public use.
Human-Centered Approaches to Supporting Fairness in AI
Vivek Krishnamurthy of University of Ottawa gave a talk titled “Human-Centered Approaches to Supporting Fairness in AI”.
The Future of AI for Social Good
Saiph Savage demonstrates how AI-based coaching on crowd-sourcing platforms can boost worker wages, efficiency, and skills while promoting fairness.
Ethics in AI: A Challenging Task
Ricardo Baeza-Yates from Institute for Experiential AI at Northeastern University gave a talk titled “Ethics in AI: A Challenging Task”.
Human-Centered Approaches to Supporting Fairness in AI
Michael Madaio from Microsoft Research discusses human-centered approaches to fairness in AI, focusing on context-driven checklists across AI design phases.
Designing Artificial Intelligence to Navigate Societal Disagreement
Michael Bernstein of Stanford introduced 'jury learning,' a method addressing bias in AI by ensuring underrepresented groups influence classifier predictions.
Maintaining fairness under distribution shift
Jessica Schrouff from Google Research explores the problem of distribution shift in healthcare AI, focusing on dermatology and electronic health records.
Is Legal AI Ethical AI?
Harvard researchers discuss the ethical challenges of legal AI, exploring how moral principles are translated into algorithms, and the risks of unintended discrimination.
Who are we listening to? Building blocks for trustworthy AI
Hillary Juma of Mozilla Foundation introduces the Common Voice project, improving voice tech accessibility by collecting 13,000 hours of data in 76 languages.
AI systems should cause no direct harm and always aim to minimize indirect harmful behavior. They should always perform as intended and remain resilient against threats and outside tampering.
Designing for Online Safety: Towards More Equitable and Trustworthy Systems
Daricia Wilkinson's study reveals 90% of Caribbean participants face online threats, from targeted ads to compromised logins, highlighting overlooked challenges.
AI systems must respect privacy by providing individuals with agency over their data and the decisions made with it. AI systems must also respect the integrity of the data they use.
Privacy and Synthetic Data: The Good, The Bad, and The Ugly
Generative models with differential privacy create synthetic datasets, but future models must ensure fairness for underrepresented groups while preserving privacy.
AI systems must be explainable and understandable. They should allow for human agency and oversight by producing outputs that are comprehensible to the average person. AI decisions should be auditable and traceable, which will ultimately instill the trust AI needs for broad acceptance.
Explainable AI for LLMs
Wojciech Samek introduces an advanced LRP method for tackling biases in large language models, offering fast, faithful input attribution and deep insights into model reasoning
Evaluating and Surgically Improving Generalization in Language Models
Aaron Mueller discusses challenges in language model generalization, covering projects on language structure, chain-of-thought reasoning, and mechanistic interpretability to…
How do we make explanations of recommendations beneficial to different users?
Explainable AI tailors explanations to systems, ensuring accuracy in personal recommendations and fairness in group content suggestions.
Value-based Engineering with IEEE 7000
Value-Based Engineering (VBE) ensures ethical AI by aligning design with prioritized values, fostering innovation, and addressing moral challenges effectively.
Mind the gap: From predictions to ML-informed decisions
Maria De-Arteaga's study on a child-welfare AI tool highlights how to balance algorithmic aversion and automation bias for effective human-AI collaboration.
Deep Neural Networks, Explanations, and Rationality
Generative Adversarial Networks can create human-like, intelligible AI explanations by pitting two systems against each other to enhance clarity and understanding.
Human-Centered Explainable AI (XAI): From Algorithms to User Experiences
Vera Liao of Microsoft Research discusses how AI explainability affects user trust, highlighting the need for designs that promote slow, careful thinking.
Responsible AI – getting the human back into the loop
Simone Stumpf outlines four principles for AI explanations: sound, complete, iterative, and user-friendly, emphasizing user involvement for improved accuracy.
Trustworthy AI
Mike Hind discusses IBM's Trustworthy AI tools: Fairness 360, Adversarial Robustness 360, and Explainability 360, plus FactSheets for full AI lifecycle transparency.
A critical look at proprietary computer vision algorithms and the data practices
Jahna Otterbacher explores biases in crowd-sourced data labeling for computer vision systems, highlighting how demographics and external events influence outcomes.
Bridging AI and HCI: Incorporating Human Values into the Development of AI Technologies
Haiyi Zhu from Carnegie Mellon University presents AI tools that balance community values in Wikipedia moderation and child protection systems.
AI systems should attempt to be both societally sustainable, by empowering society and democracy, and environmentally sustainable, by reducing the amount of power required to train and run these systems.
AI systems should be developed and deployed through consultation and collaboration in order to create true accountability. The long-term effects of an AI application must be understandable by all stakeholders, and those stakeholders must be empowered to act if any proposed change produces adverse effects.
How to Unknow the Uncertainties in Data Science
Michael Muller of IBM Research and Angelika Strohmayer of Northumbria School of Design gave a talk titled “How to Unknow the Uncertainties in Data Science”.
AI Governance for 2024: Telco Edition
Kevin Schawinski discusses AI regulation, focusing on the EU AI Act's impact on processes, governance, and compliance, especially for critical industries like telecom, and how…
The Property of Intelligence: When AI meets IP
Ben Maling discusses AI and intellectual property, covering IP issues in generative AI, open-source paradoxes, patents, and the EU AI Act's potential impact on these dynamics.
Discover more
Biographies
Location
Service
Awards