Papers of the day   All papers

Model Cards for Model Reporting


Nov 21 2019 Jeff Dean

Great to see more transparent and interpretable ML models & tools being made available via @GCPcloud. Particularly excited that Model Cards, which started as a @GoogleAI research project by @mmitchell_ai et al. are part of this (
4 replies, 512 likes

Dec 16 2019 Thomas Wolf

Current status: adding model cards for the 63 (!) pretrained models in 🤗 Transformers library 👉 more transparent reporting on models' capabilities, limitations, uses... Read (@mmitchell_ai @timnitGebru et al.) for details. Example (from this paper)👇
4 replies, 290 likes

Oct 12 2018 Timnit Gebru

@mmitchell_ai did not waste any time putting me to work! Excited for my first of hopefully many collaborations with her and many others. Model Cards for Model Performance is analogous to Datasheets for Datasets :)
4 replies, 148 likes

Oct 28 2019 DPManchee

@jovialjoy @StanfordHAI How to identify bias in models / datasets ? Adopt ideas from Mitchell et al (2018), "Model Cards For Model Reporting", ? Discussed more here at a @stanfordnlp lecture,
0 replies, 35 likes

Dec 03 2019 HotComputerScience

Most popular computer science paper of the day: "Model Cards for Model Reporting"
1 replies, 28 likes

Sep 23 2019 Arvind Narayanan

Oops, wrong link to "Model Cards for Model Reporting" earlier in the thread (link goes to "Datasheets for Datasets" instead). Correct link: I meant to say that Datasheets for Datasets is *also* a great paper on how to do ML in a more responsible way :)
1 replies, 25 likes

Nov 22 2019 Tracy Pizzo Frey

*Tracy stay professional* *Tracy you are an adult* *Tracy take a deep bre--* OMG I AM SO EXCITED I CAN'T HOLD IT IN THIS JUST HAPPENED
1 replies, 14 likes

Nov 24 2019 Marcus East

One of the barriers to the adoption of AI has been the opaqueness of its inner workings — this announcement will help to improve transparency and interpretability, critical in decisions connected with financial services, law enforcement and medicine. #google #googlecloud #ai #ml
0 replies, 7 likes

Nov 20 2019 Larissa Schiavo

Idea (for the graphic design gifted): model cards for trained machine learning models (see paper)... ... made into physical cards, designed like baseball cards. AI people give them to other AI people and trade them.
2 replies, 7 likes

Nov 23 2019 Hugo Bowne-Anderson

I’m pretty excited about this, particularly the incorporation of Model Cards for Model Reporting, which is one of my fave papers of the ~past year (thanks @mmitchell_ai @timnitGebru et al; & thanks to @mikeloukides for introducing me to the paper).
0 replies, 7 likes

Nov 25 2019 Joseph Futoma

Great to see Model Cards ( picking up steam at Google! h/t @jackclarkSF from his always-informative ImportAI newsletter! Looking fwd to when this is standard & reqd across field, like @TRIPODStatement (@GSCollins have u seen MCs?)
0 replies, 6 likes

Aug 03 2019 andrea panizza

Learnt today about Datasheets for Datasets by @timnitGebru et al. and Model Cards for Model Performance by @mmitchell_ai et al. both sound like very good ideas. Are they used in practice?
0 replies, 4 likes

Dec 03 2019 Data Nutrition Project

.@mmitchell_ai @DrLucyWasserman @timnitGebru et al. advocated for a framework of transparent model reporting to accompany trained ML models as a step towards responsible AI works in their paper We're glad to be referenced in your work!💫
0 replies, 4 likes

Nov 23 2019 mauro_iker

Model Cards: Explainable model reporting to document any trained #MachineLearning model #XAI #ExplainableAI #ArtificialIntelligence #AI #Standards #DeepLearning #DL #NeuralNetworks #NN #ComputerVision #NLP
0 replies, 1 likes

Feb 25 2020 Marcel Fröhlich

Besides data transparency, #GDPR also requires algorithmic transparency for automated decisions. Most of that is a metadata exercise. The missing piece is usually the actual trained model, in case of #ML #algorithms. Here is a way to tackle that, too:
0 replies, 1 likes

Apr 26 2019 Mia Dand

@lizjosullivan A1. Here is a great example of Model Cards for Model Reporting by @timnitgebru Margaret Mitchell and others #AIEthics #TwitterChat #Transparency #Explainability
0 replies, 1 likes