ServiceNow Research

Trustworthiness

Explainable, Sensible and Virtuous Workplace Chatbots
We outline three research directions towards the practical implementation of explainable, sensible and virtuous chatbots for the …
Montreal AI Symposium (MAIS),  2022.
Cite
Examining Responsibility and Deliberation in AI Impact Statements and Ethics Reviews
Examining Responsibility and Deliberation in AI Impact Statements and Ethics Reviews
The artificial intelligence research community is continuing to grapple with the ethics of their work by encouraging researchers to …
AAAI/ACM Conference on AI, Ethics, and Society (AIES),  2022.
PDF Cite
The Dynamics of Functional Diversity throughout Neural Network Training
Deep ensembles offer consistent performance gains, both in terms of reduced generalization error and improved predictive uncertainty …
Conference on Neural Information Processing Systems (NeurIPS),  2021.
PDF Cite
Pruning Neural Networks at Initialization: Why are We Missing the Mark?
Pruning Neural Networks at Initialization: Why are We Missing the Mark?
Recent work has explored the possibility of pruning neural networks at initialization. We assess proposals for doing so: SNIP (Lee et …
International Conference on Learning Representations (ICLR),  2021.
PDF Cite
On the role of data in PAC-Bayes bounds
On the role of data in PAC-Bayes bounds
The dominant term in PAC-Bayes bounds is often the Kullback–Leibler divergence between the posterior and prior. For so-called …
International Conference on Artificial Intelligence and Statistics (AISTATS),  2021.
PDF Cite Code
Using Harms and Benefits to Ground Practical AI Fairness Assessments in Finance
ACM Conference on Fairness, Accountability, and Transparency (FAccT),  2021.
Cite
An empirical study of loss landscape geometry and evolution of the data-dependent Neural Tangent Kernel
Conference on Neural Information Processing Systems (NeurIPS),  2020.
PDF Cite
Like A Researcher Stating Broader Impact for the Very First Time
Like A Researcher Stating Broader Impact for the Very First Time
In requiring that a statement of broader impact accompany all submissions for this year’s conference, the NeurIPS program chairs …
Workshop at the Neural Information Processing Systems (NeurIPS),  2020.
PDF Cite
Pruning Neural Networks at Initialization: Why Are We Missing the Mark?
Pruning Neural Networks at Initialization: Why Are We Missing the Mark?
Recent work has explored the possibility of pruning neural networks at initialization. We assess proposals for doing so: SNIP (Lee et …
Workshop at the Neural Information Processing Systems (NeurIPS),  2020.
PDF Cite
Sharpened Generalization Bounds based on Conditional Mutual Information and an Application to Noisy-Gradient Iterative Algorithms
Sharpened Generalization Bounds based on Conditional Mutual Information and an Application to Noisy-Gradient Iterative Algorithms
The information-theoretic framework of Russo and J. Zou (2016) and Xu and Raginsky (2017) provides bounds on the generalization error …
Conference on Neural Information Processing Systems (NeurIPS),  2020.
PDF Cite