44 Hits in 2.9 sec

Scrutable Feature Sets for Stance Classification

Angrosh Mandya, Advaith Siddharthan, Adam Wyner
2016 Proceedings of the Third Workshop on Argument Mining (ArgMining2016)  
The main advantage of our feature set is that it is scrutable: The reasons for a classification can be explained to a human user in natural language.  ...  This paper describes and evaluates a novel feature set for stance classification of argumentative texts; i.e. deciding whether a post by a user is for or against the issue being debated.  ...  Full dependency model Our proposed feature set for stance classification using a distributional lexical model, stance bearing terminology and points was designed to be scrutable, but therefore made use  ... 
doi:10.18653/v1/w16-2807 dblp:conf/acl/MandyaSW16 fatcat:zbdjepiopffbnfbopvtj6putjm

Explaining Explanations: An Overview of Interpretability of Machine Learning [article]

Leilani H. Gilpin, David Bau, Ben Z. Yuan, Ayesha Bajwa, Michael Specter, Lalana Kagal
2019 arXiv   pre-print
We discuss why current approaches to explanatory methods especially for deep neural networks are insufficient.  ...  Finally, based on our survey, we conclude with suggested future research directions for explanatory artificial intelligence.  ...  The authors also wish to express their appreciation for Jonathan Frankle for sharing his insightful feedback on earlier versions of the manuscript.  ... 
arXiv:1806.00069v3 fatcat:zegbomvrrredxazh2t7z2og4ju

Mitigating Bias in Algorithmic Systems - A Fish-Eye View

Kalia Orphanou, Jahna Otterbacher, Styliani Kleanthous, Khuyagbaatar Batsuren, Fausto Giunchiglia, Veronika Bogina, Avital Shulner-Tal, Alan Hartman, Tsvi Kuflik
2021 Zenodo  
space based on the sensitive feature set.  ...  [173] use a multi-objective optimization algorithm for feature construction.  ... 
doi:10.5281/zenodo.6240582 fatcat:vftoi4woebhrrp5tlmkclabgf4

A giant with feet of clay: on the validity of the data that feed machine learning in medicine [article]

Federico Cabitza, Davide Ciucci, Raffaele Rasoini
2018 arXiv   pre-print
In the case of medical classification (for both diagnostic or prognostic aims) the above "experience" is a set of cases that have been already represented properly and classified "correctly" according  ...  A ML-DSS that has learned the uncertain (i.e., right for a standard, wrong for another) mapping between the patient's features and one single diagnosis will propose its advice within a dangerous "close-world  ... 
arXiv:1706.06838v3 fatcat:xmfzemujszb43gz4a3h3phswtm

As if sand were stone. New concepts and metrics to probe the ground on which to build trustable AI

Federico Cabitza, Andrea Campagner, Luca Maria Sconfienza
2020 BMC Medical Informatics and Decision Making  
We argue that the proposed metrics are feasible for application in real-world settings for the continuous development of trustable and interpretable MAI systems.  ...  We define a set of quality dimensions and related metrics: representativeness (are the available data representative of its reference population?)  ...  Thus, each case is labeled with a unique value for the target feature.  ... 
doi:10.1186/s12911-020-01224-9 pmid:32917183 fatcat:spt7thb24banvpmqodcsfdwv2q

Special issue of synthese on new directions in the epistemology of modality: introduction

Antonella Mallozzi
2019 Synthese  
Thanks also to Otávio Bueno and Wiebe van Der Hoek for their kind editorial guidance.  ...  Acknowledgements Thanks to Nathan Bice, Paul Boghossian, Michael Devitt, David Papineau, Jonathan Schaffer, Anand Vaidya, and Michael Wallner for helpful conversation and feedback.  ...  The resulting theories, while positively essentialist, look hybrid according to traditional classifications (e.g.  ... 
doi:10.1007/s11229-019-02358-8 fatcat:no57rs5khvb5hckmw3lnjc2voy

Explanation in Human-AI Systems: A Literature Meta-Review, Synopsis of Key Ideas and Publications, and Bibliography for Explainable AI [article]

Shane T. Mueller, Robert R. Hoffman, William Clancey, Abigail Emrey, Gary Klein
2019 arXiv   pre-print
This is an integrative review that address the question, "What makes for a good explanation?" with reference to AI systems. Pertinent literatures are vast. Thus, this review is necessarily selective.  ...  The test messages were designed such that only one feature was important to the classification, such as an obvious feature or a subtle feature.  ...  Performance (Task or Evaluation Measure) Participants could add features to the classification scheme, remove features or adjust features' importance.  ... 
arXiv:1902.01876v1 fatcat:glklvdwc2rbdzgteew4ttfjz2q

Learning from Artificial Intelligence's Previous Awakenings: The History of Expert Systems

David C. Brock
2018 The AI Magazine  
The whole tradition of knowledge engineering evolved into methods for acquiring features, relationships, and parameters.  ...  For Mahoney, a defining activity of these various communities of computing was creating software that, for him, constituted the modeling of certain features of the physical or social world.  ... 
doi:10.1609/aimag.v39i3.2809 fatcat:u4ofx2p2kng2ncon3xpafwl3q4

Disembodiment and Identity in Literary Depictions of Epilepsy Surgery

Maria Vaccarella
2015 Literature and medicine  
The hospital panels set the atmosphere for Professor T.'  ...  41 At this point, Jean-Christophe's disembodied brain enlarges in the panels and becomes a mere setting for yet another of the surgeon's medical adventures.  ... 
doi:10.1353/lm.2015.0004 pmid:26095838 fatcat:74e5obqvlvfrjldusk5bmtielm

Influence Tuning: Demoting Spurious Correlations via Instance Attribution and Instance-Driven Updates

Xiaochuang Han, Yulia Tsvetkov
2021 Findings of the Association for Computational Linguistics: EMNLP 2021   unpublished
For One notable feature of influence tuning is that it text classification, the prediction head is just a lin- is designed to help deconfounding NLP models ear projection layer  ...  However, as the spurious attribute To interpret each classification decision via in- should not influence the model’s decision process, stance attribution, we randomly sample a few ex-  ... 
doi:10.18653/v1/2021.findings-emnlp.374 fatcat:4set4sfpqba4liycipvs7wsc6q

Impact of Implicit and Explicit Affective Labeling on a Recommender System's Performance [chapter]

Marko Tkalčič, Ante Odić, Andrej Košir, Jurij Franc Tasič
2012 Lecture Notes in Computer Science  
This talk will give a sample of key concepts and results from these areas, showing how they suggest new research issues and design ideas for those who work on recommender systems.  ...  (iii) classification.  ...  To give the user the chance to correct the system (scrutability) and to deliver effective recommendations is important for recommendation systems in general.  ... 
doi:10.1007/978-3-642-28509-7_32 fatcat:7t5x5uxsgbfeniqp6qhqpynpl4

Explainable Software for Cyber-Physical Systems (ES4CPS): Report from the GI Dagstuhl Seminar 19023, January 06-11 2019, Schloss Dagstuhl [article]

Joel Greenyer, Malte Lochau, Thomas Vogel
2019 arXiv   pre-print
This report documents the program and results of the GI-Dagstuhl Seminar 19023 on Explainable Software for Cyber-Physical Systems (ES4CPS).  ...  The seminar was concerned with the challenge that for future Cyber-Physical Systems (CPS), it will become increasingly relevant to explain their behavior (past, current, and future behavior, why a certain  ...  Acknowledgments We would like to thank the Gesellschaft for Informatik e.V. (GI, German Society for Informatics) and Schloss Dagstuhl for partially funding this seminar.  ... 
arXiv:1904.11851v1 fatcat:r2r5ac6hmnel3f6weiynjwunni

The chromatin bodies of bacteria

1956 Bacteriological reviews  
Significant features of these methspaces against the pink cytoplasm. ods are set out in the following discussion: The influence of the composition of the medium Iron alum hematoxylin (IAH).  ...  Its essential feature is the cells possess organelles which divide in a regular duplication and orderly segregation of the whole manner, are cag with DNoA but lack visible. . r charged set of chromosome  ... 
pmid:13403844 pmcid:PMC180865 fatcat:khguhn5rtbczxizv42b5lao2fm

Towards a framework for certification of reliable autonomous systems

Michael Fisher, Viviana Mascardi, Kristin Yvonne Rozier, Bernd-Holger Schlingloff, Michael Winikoff, Neil Yorke-Smith
2020 Autonomous Agents and Multi-Agent Systems  
However, regulators grapple with how to deal with autonomous systems, for example how could we certify an Unmanned Aerial System for autonomous use in civilian airspace?  ...  Thanks to Simone Ancona for the drawings in Sect. 1.  ...  Setting up a proof correctly is a difficult activity even for trained experts.  ... 
doi:10.1007/s10458-020-09487-2 fatcat:gd6urk3nbrcwxkpffdtdpa5xha

Towards a Framework for Certification of Reliable Autonomous Systems [article]

Michael Fisher, Viviana Mascardi, Kristin Yvonne Rozier, Bernd-Holger Schlingloff, Michael Winikoff, Neil Yorke-Smith
2020 arXiv   pre-print
However, regulators grapple with how to deal with autonomous systems, for example how could we certify an Unmanned Aerial System for autonomous use in civilian airspace?  ...  Thanks to Simone Ancona for the drawings in Section 1.  ...  Work of Fisher supported by the Royal Academy of Engineering and, in part, through UKRI "Robots for a Safer World" Hubs EP/R026092, EP/R026084, and EP/R026173.  ... 
arXiv:2001.09124v1 fatcat:ocmnn2ugzfeatfc56ucwgz2aku
« Previous Showing results 1 — 15 out of 44 results