Happy to share that version 2.0 of our Explanation Ontology has been accepted as a journal paper to the Special Issue in XAI at the Semantic Web Journal (SWJ). Link to paper: https://t.co/okrriZWbZz 1/6

Happy to share that I am now a PhD candidate and I am looking forward to a defense next year on my dissertation research titled, "An Ontology-Enabled Approach For User-Centered and Knowledge-Enabled Explanations of AI Systems" #phdlife @csatrpi

A really nice article in the WSJ today about the various use cases of AI in hospitals across the US: https://t.co/6oCsJZDzH3. It helps to both understand the trends of application of AI in healthcare and identify where we can improve. Thanks for sharing @dangruen.

Paper finally out in a journal! https://t.co/ZbRnOAuiXm @shruthichari, @oshaniws, @mo_ghalwash, @dangruen,@pchakrabt1,@dlmcguinness

Many congratulations to my advisor, Prof. Deborah McGuinness as being elected as a AAAI Fellow 2023!

Thanks to the efforts of all my collaborators and co-authors from @csatrpi and @IBMResearch - @oshaniws , @mo_ghalwash, @dangruen, Sola Shirai, @pchakrabt1, @jeriscience and @dlmcguinness. 6/6

You can download our ontology and explore our resource from our website: https://t.co/g5y0jaxqBS. Looking forward to feedback on using the EO or queries about applying it to your own use cases. 5/6

Here is a list of explanation types we support. Encoding the explanations using our model, also allows them to be inferred into one or more of the explanation types we support in the EO. 4/6

Below is an image of our updated model of the EO. The EO is a resource to model explanations based on their dependencies from the user, interface and system attributes. 3/6

In V2 of the EO, we support more explanation types taking our tally to 15 and also introduce better compatibility with explainer models mainly from the IBM AIX-360 toolkit. 2/6

.@Stanford’s PubMedGPT LLM passes US Medical Licensing Exam (MedQA-USMLE) with more than 50% correct answers.

Nice thread on the shortcomings of LLMs and why we must build guards around them.

Never had a failed macOS upgrade before Ventura and it is painful to get your Mac to start again -.- Maybe finally a time where I appreciate more transparent options aka Linux.

I can't believe in 2022 we're still sending reporters out in Hurricanes like this

Hope such efforts bring to light the need to focus more on reporting results and releasing models more often outside of general-purpose domains. Also, hope that AI/ML papers start asking for more rigorous and illustrative analyses outside of just quantitative numbers. 2/2

Happy to see an application of language models (LMs) in a high-stakes setting like the government. I like how the post author says it is to 'show, not tell' in this thread. Working on LMs, I was frustrated with how little there is outside of large corpora like Wikipedia. 1/2

Unpopular Opinion: Almost all academic papers do a decent job of *what* they found, many do pretty well to describe *how* they did it, very few do a stellar job of *why* they are doing it and *why* the community needs it. @PhDVoice @OpenAcademics #AcademicTwitter #research

A common point raised by ML reviewers is that a method is too simple or is made of existing parts. But simplicity is a strength, not a weakness. People are much more likely to adopt simple methods, and simple ones are also typically more interpretable and intuitive. 1/2

Very excited to introduce, OpenXAI - https://t.co/gSPMNgdlqq, an open-source framework we have been building for the past year to evaluate/benchmark the faithfulness, stability, fairness of post hoc explanation methods using easy-to-use API & just a few lines of code. [1/N] #XAI

A comprehensive thread of XAI papers presented at this year’s XAI workshop at #IJCAI2022. Do give them a read if you are doing researching in the XAI space or are looking to improve explainability of your AI methods.