Future Direction

[bwang514]

Hey collaborators, I think it would be better to define the scope of this repository to human-ai interaction. My original idea was to create a curation for these resources, but as @fanglinchen suggested, this could go beyond just a curation to potentially creating new knowledge based on the digestion of these papers. I want to open this thread for everyone to throw out ideas/opinions about how can we strengthen our repo! Feel free to comment!

@minsukchang @jritch @luzc08 @joe32140 @mingming-fan @fanglinchen

[mingming-fan]

Thanks for sharing Bryan! What does “potentially creating new knowledge based on digestion of these papers” encompass? Are you envisioning discussions among us about potential ideas that might be interested to work on?

Cheers,

Mingming

From: Bryan Wang notifications@github.com Sent: Wednesday, May 15, 2019 12:57 PM To: bwang514/awesome-HAI awesome-HAI@noreply.github.com Cc: Mingming Fan mingming.fan@mail.utoronto.ca; Mention mention@noreply.github.com Subject: [bwang514/awesome-HAI] Future direction of this repo (#11)

Hey collaborators, I think it would be better to define the scope of this repository to human-ai interaction. My original idea was to create a curation for these resources, but as @fanglinchen https://github.com/fanglinchen suggested, this could go beyond just a curation to potentially creating new knowledge based on the digestion of these papers. I want to open this thread for everyone to throw out ideas/opinions about how can we strengthen our repo! Feel free to comment!

@minsukchang https://github.com/minsukchang @jritch https://github.com/jritch @luzc08 https://github.com/luzc08 @joe32140 https://github.com/joe32140 @mingming-fan https://github.com/mingming-fan @fanglinchen https://github.com/fanglinchen

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/bwang514/awesome-HAI/issues/11?email_source=notifications&email_token=AHYF5YERKAHB7J7FS3PFEEDPVQ6HDA5CNFSM4HNFGKP2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4GT7HFMQ , or mute the thread https://github.com/notifications/unsubscribe-auth/AHYF5YBXTVOW6RNETZEKVN3PVQ6HDANCNFSM4HNFGKPQ .

[bwang514]

Hi @mingming-fan , potential research ideas could also be useful! But what I meant was perhaps we could find an underlying trend of how researchers deploy AI into interfaces or more focused subsets like the use of chatbots in HCI, what are the capabilities and challenges of current state-of-the-art research? It would be more like a literature review that points out general concepts of what is Human-AI Interaction at this moment, and maybe eventually it could turn into a paper or something at next year's CHI. I don't really have a concrete idea on how to do this, so I hope there could be some interesting discussions here!

[fanglinchen]

@bwang514 agreed.

If we think of the current format of the curation, although we have put together AI-relevant papers in the HCI field, we don't have too many digestible forms of knowledge to the first-time viewers. One thought would be to come up with a taxonomy of the existing contents and present the papers according to the taxonomy.

I think it would be also impactful if we can identify the key areas of concerns and research interests in this rising domain of HAI, through analyzing the papers published over the years.

[tcboy88]

I personally do not like combining HCI-DL with HAI. Previously it was more tech focussed and now it becomes too broad, with all the theory and guidelines stuffs. While I am not a contributor on this repo, I prefer if you guys can separate HCI-DL vs. HAI, revert back to just HCI-DL and maybe create another repo just for HAI?

[minsukchang]

I don't know if there needs to be a distinction between HCI-DL and HAI. HCI has always been about making technology more accessible to people, and AI (surprise!) was one of those technologies for at least 30 years. But we could separate a section for "guidelines", because they're a different breed.

One thing this curated list and a bit of a taxonomy work (as suggested above) could reveal is what user problems we can tackle with currently available "neural" techniques in the differentiable world.

and btw, we're missing literature from ML venues that tackle HCI problems too.

I personally do not like combining HCI-DL with HAI. Previously it was more tech focussed and now it becomes too broad, with all the theory and guidelines stuffs. While I am not a contributor on this repo, I prefer if you guys can separate HCI-DL vs. HAI, revert back to just HCI-DL and maybe create another repo just for HAI?

[minsukchang]

To follow up, what are we going to do with all the work coming from venues like HCOMP, FAT/ML, AAAI, NeurIPS, ICML, EDM, L@S, etc. ?

[bwang514]

Good points Minsuk! I think we can just include them if related. As the number of papers increases, we could categorize them by the type of problem they tackled, instead of by the venues.