From OntologPSMW

Jump to: navigation, search
[ ]
    (1)
Session Communique Development
Duration 1 hour60 minute
3,600 second
0.0417 day
Date/Time May 01 2019 16:00 GMT
9:00am PST/12:00pm EST
5:00pm GMT/6:00pm CET
Convener Ken Baclawski

Contents

Ontology Summit 2019 Communique Development Session 2     (2)

Agenda     (2A)

Discussion of the Track Summaries     (2A1)

Conference Call Information     (2B)

Attendees     (2C)

Proceedings     (2D)

Panel Questions     (2D1)

  • Can computer-based explanations be made as informative and useful as an explanation from a well-informed person?     (2D1A)
    • Are there disconnects among researchers, industry and media, or between users and investors with respect to what constitutes an acceptable or successful explanation?     (2D1A1)
    • Is there general ignorance about what is already possible vs what is well beyond current capabilities?     (2D1A2)

Discussion     (2D2)

[12:07] Ram D Sriram: @David: We need to talk about the Medical XAI. Let me know of a good time tomorrow morning or Friday morning. Please send me email (sriram@nist.gov with some times and a tel. number)     (2D2A)

[12:08] Gary: See some input on the Communique under the Resources part of this session. http://ontologforum.org/index.php?title=ConferenceCall_2019_05_01&oldid=25653#hid2E     (2D2C)

[12:20] RaviSharma: Ken - Background precedes the track summaries     (2D2D)

[12:20] RaviSharma: Gary thanks for providing new communique outline     (2D2E)

[12:32] John Sowa: Time control is very important. Without strict time control, later speakers get shut out or pushed beyond prime time.     (2D2F)

[12:32] RaviSharma: Ram and Ken others - Sargur Srihari, Mike Gruninger, Giedrius, Derek, Bill Clancey, were some speakers suggested     (2D2G)

[12:33] RaviSharma: Ravi - feels we should increase the final synthesis time on 8th May to include less covered topics such as :     (2D2H)

[12:34] RaviSharma: Cross track Integration     (2D2I)

[12:34] RaviSharma: Role of Explanations in AI and Ontology solutions     (2D2J)

[12:35] RaviSharma: Future Directions in Maturity and use of Explanations and allied topics     (2D2K)

[12:35] RaviSharma: Ken- overlap among tracks     (2D2L)

[12:36] RaviSharma: Ravi - usually Ken and others help achieve this in the language of the Communique     (2D2M)

[12:36] John Sowa: To get cross-track integration, the people who are presenting each track must share their slides prior to any presentation.     (2D2N)

[12:37] RaviSharma: I agree with John     (2D2O)

[12:37] RaviSharma: Let us set deadline for Synthesis and draft Communique inputs to 4th May 3 days from now at the latest.     (2D2P)

[12:38] Ken Baclawski: Each track should use the same outline: Introduction/Background, Summary using slides from the speaker presentations, Cross-Track Integration, Opportunities & Challenges, Conclusions.     (2D2Q)

[12:43] RaviSharma: Bobbin and Terry and Bruce may be involved     (2D2R)

[12:49] RaviSharma: Commonsense and Narrative on first day     (2D2S)

[12:49] RaviSharma: other 3 on second day     (2D2T)

[12:53] RaviSharma: 8th may is wed     (2D2U)

[12:53] RaviSharma: sample Questions for panel     (2D2V)

[12:53] Ram D Sriram: What are potential panel questions?     (2D2W)

[12:56] RaviSharma: at least panelists should have a place to read draft communique before they come     (2D2X)

[12:57] RaviSharma: janet suggests teasers or challenge for panelists     (2D2Y)

[12:58] RaviSharma: My Questions are similar to John as to how well are we having HCI with Explanations added than without it, by context and agreed shared Vocabularies     (2D2Z)

[13:02] janet singer: Challenge question following John: When can we expect machine-to-human explanations to achieve the level of human-to-human explanations? What should be done in the meantime to mitigate the quality gap between the two?     (2D2AA)

[13:04] RaviSharma: John emphasized context, explanation and fixed scripts     (2D2AB)

[13:06] RaviSharma: I would also like to ask panel on effectiveness and value add metrics when you use Explanations and when you do not.     (2D2AC)

[13:06] janet singer: Challenge question per Mike: What should an ontology be like to support explanations?     (2D2AD)

[13:06] RaviSharma: Also context, does it fill that gap instead of explanations somewhat?     (2D2AE)

[13:07] RaviSharma: yes some prelim Qs upfront are desirable     (2D2AF)

[13:08] John Sowa: The list of questions could be updated during the next two weeks.     (2D2AG)

[13:09] John Sowa: The first list could be updated every few days, as people get more "bright ideas".     (2D2AH)

[13:09] Gary: Q...I like the idea of a panel Q like, "What knowledge and reasoning do we need regarding ontologies to support Explanation?" But what type of answer do we have or are developing in our Communique?     (2D2AI)

[13:10] John Sowa: I have to sign off at 1:15     (2D2AJ)

[13:14] John Sowa: Question: How can we make computer explanations more flexible and human like?     (2D2AK)

[13:16] John Sowa: I have to sign off     (2D2AL)

[13:24] Ram D Sriram: Have to go to another meeting. Signing off     (2D2AM)

Resources     (2E)

Here is a working abstract based on what we did in past years.     (2E1)

There are many disciplined involved in understanding explanation and implementing systems which provide explanations. Among these are artificial intelligence, semantic and ontological science, machine learning, and linguisitics. Semantic models, ontologies, neural net models with their algorithms along with knowledge representation provide important areas of work. The Ontology Summit 2019 explored, identified and articulated the relationships among these areas and across domains such as medicine, finance and narration. In 2019, as in prior years, an important aspect of Ontolog Forum advocacy has been to bring ontology science and engineering into the mainstream. In particular we continue to endeavore to synthesze a conversational toolkit from the Ontology Summit sessions in order to facilitate discussion and knowledge sharing amongst stakeholders concerned with the topic. Our findings are supported and illustrated with examples from various domains of interest. The results were captured in the form of this Communique, with ´ expanded supporting material provided on the Ontolog website. Keywords: artificial intelligence; machine learning; explanation; ontologies, XAI, deep learning     (2E2)

Here is a working outline for the 2019 O-Summit Communique     (2E3)

1. Introduction We are currently witnessing increasingly widespread applications of Artificial Intelligence (AI), which deals with intelligent explanation. DAPRA is sponsoring, X is building …     (2E4)

How we conducted the Summit – tracks...synthesis….     (2E5)

 Logo diagram??

2 Track Summaries 2.1 Overview Session Each session to provide an abstract of their summary     (2E6)

3. Background     (2E7)

   We had some history of explanation from Ram, Gary, John that can be used here.

4 Opportunities and Challenges A great many techniques have been developed for explanations that make use of proofs, maching learning, commonsense reasoning, ontologies and (TBD) This section is an attempt to survey some of the problems and opportunities of the relationships among these.     (2E8)

the driver from XAI to provide better explanations

Infusion of commonsense and NL understanding Knowledge representation…...     (2E10)


5 Conclusion The 2019 Ontology Summit examined a wide range of issues, opportunities, challenges and future prospects for automated and understood explanations with interconnections among and between efforts in AI, ML, CSK &reasoning and ontological science.     (2E11)

We learned that 

We discussed the problem of understanding….     (2E12)

We noted that AI in general, and ML in particular, is 

We revisited the issue of We examined the problem of We explored the processes that     (2E13)

We showed how 

6 Acknowledgments Certain commercial software systems are identified in this paper…... We wish to acknowledge the support of the ontology community, and the support of NIST who generously contributed the use of their facilities for the Ontology Summit Symposium.     (2E14)

References

Previous Meetings     (2F)


Next Meetings     (2G)