Thanks Denny! I appreciate this, and your thoughtfulness as always. Thanks for starting an explicit discussion. I think our field (meaning, computer science & internet projects broadly) often assumes that these questions will just arise in context, or get solved as we go, but they rarely do. So making ethics a focus from the start is crucial. There may not be good "solutions"! But inviting lots of people in to talk about scenarios etc I think will make a much stronger and innovative project in the end.
all best, Phoebe On Fri, Jul 3, 2020 at 7:54 PM Denny Vrandečić <vrande...@gmail.com> wrote: > Thank you Phoebe for the congratulations, and thank you for the > considerations, and I agree with them. > > On wiki, Denis Barthel and Ryan Kaldari already raised a number of the > concerns - although not all - that you are raising here. > > https://meta.wikimedia.org/wiki/Talk:Abstract_Wikipedia#Kaldari's_concerns > > > https://meta.wikimedia.org/wiki/Talk:Abstract_Wikipedia#Revisiting_and_twisting_a_Kaldari_concern > > And I tried to answer some of them there. But I do agree that this is not > sufficient, and I would love to be able to ensure that the ethical concerns > you raise do not get lost, and are sufficiently represented when we do the > project. > > Alas, I am also a bit at a loss about how to ensure that. Yes, for some of > them, we have a pretty good, albeit developing understanding of how to > cover them, particularly around ethics and AI and about bias and ethical > data. Also I think doing several best / worst case scenario exercises at > relevant points is a great idea. The one thing that troubles me most, > though, is how to ensure that in the new communities that we will foster > the representation among contributors is indeed more representative of the > diversity in the world. To the best of my knowledge, we have no answers for > that - and I would very much want to learn about this. > > So, here's what I can promise - among the many topics that we need to > discuss while we are ramping up the project, I will also start an explicit > discussion on how to make sure that ethical considerations are sufficiently > represented during the development of the project. I obviously cannot > promise that we will successfully avoid all ethical pitfalls - but I can > promise that I will do my best to do so. > > It is, in the end, ethical concerns that motivated me, and some of them > are discussed and described here: > > https://wikipedia20.pubpub.org/pub/vyf7ksah/release/6 > > It is this motivation of allowing more people to share in more knowledge > in more languages which drives me. > > I hope you'll join us on the new list and keep an eye on what we're doing. > Your voice would be very appreciated. > > https://lists.wikimedia.org/mailman/listinfo/abstract-wikipedia > > Thank you! > Denny > > > > On Fri, Jul 3, 2020 at 11:07 AM phoebe ayers <phoebe.w...@gmail.com> > wrote: > >> Congratulations to everyone, this is exciting. >> >> It is also very exciting that we have an almost-unprecedented opportunity >> to build a new project that is fully informed by both lessons from own past >> projects, as well as from the rapidly developing field of ethics in >> computer science and AI. From our own past projects, we have learned (among >> other things) that pitfalls could include contributor recruitment; >> continued maintenance; transparency and accessibility of the UI; unclear >> provenance of data; that many communities want a say in how they are >> represented online, but often don't have one; and that the biases and >> systemic biases of the world are reflected in who contributes, what sources >> they use, and what areas of focus are. We have also learned that our >> relationship with reusers, particularly around structured data that is >> highly valued by commercial entities, is poorly defined and tenuous. From >> the movement to build more ethical AI systems, we've learned (among other >> things) that flawed model assumptions can result in unpredictable and often >> deeply harmful downstream outcomes; that most sources of data are not >> transparent in their limitations or provenance; and that incorporating the >> concerns of people affected by systems can result in less biased data and >> outcomes. >> >> These are hard problems, and they are not problems that have obvious, >> one-size-fits-all solutions. But we do have an obligation I think to >> consider these issues front and center in this new project that we are >> building. It concerns me that, at least in the high-level project proposals >> I've seen (I haven't been tracking this closely, and haven't read the >> academic papers) I have not yet seen discussions of ethical data, or how we >> might think about identifying bias, or even how to recruit contributors and >> the impact on existing contributors. >> >> In my day job, I am the librarian for a pretty well-known computer >> science school.[1] There, the recent movement to consider ethics in the >> computer science curriculum, and in the systems that our computer >> scientists build, is being discussed now at all levels of the university -- >> but is being led primarily by students who recognize that they have an >> obligation, as the next generation of engineers, to help build better >> systems for a better world. Meanwhile, as practitioners who build systems >> at Wikimedia, we consider ourselves part of a small group of influential >> organizations that is "making the internet not suck" -- we believe in >> openness, in community, and in making sure that everyone in the world has >> access to knowledge, in their own language; we believe in an aspirational >> better world. As a part of this mission, we must take questions of ethics >> seriously -- and we do. We have collectively spent thousands of hours >> trying to expand our contributor base; thinking about systemic bias; >> thinking about sources and provenance; trying to open up copyright to make >> knowledge accessible; working with communities on indigenous knowledge; >> building UIs that are easier to contribute to. These are all efforts >> related to our ethics and values. With our new projects, we can set >> precedent. We can explore the problems that we face today on Wikipedia, >> Wikidata, and Commons and consider not just how to avoid them but how to >> build a better project. We can do this in a multilingual context with >> perspectives from volunteers and staff around the world, in a way that >> almost no other projects online -- certainly no single university or >> research group -- can. We can, without much legacy infrastructure to hamper >> us, spin out worst-case and best case scenarios, ask questions about our >> data and who might participate, think about downstream consequences. And >> *that* is truly exciting. >> >> best, >> -- Phoebe >> >> >> [1] >> https://en.wikipedia.org/wiki/MIT_Computer_Science_and_Artificial_Intelligence_Laboratory >> >> >> >> On Thu, Jul 2, 2020 at 12:04 PM Katherine Maher <kma...@wikimedia.org> >> wrote: >> >>> (A translatable version of this announcement can be found on Meta [1]) >>> >>> Hi all, >>> >>> It is my honor to introduce Abstract Wikipedia [1], a new project that >>> has been unanimously approved by the Wikimedia Foundation Board of >>> Trustees. Abstract Wikipedia proposes a new way to generate baseline >>> encyclopedic content in a multilingual fashion, allowing more contributors >>> and more readers to share more knowledge in more languages. It is an >>> approach that aims to make cross-lingual cooperation easier on our >>> projects, increase the sustainability of our movement through expanding >>> access to participation, improve the user experience for readers of all >>> languages, and innovate in free knowledge by connecting some of the >>> strengths of our movement to create something new. >>> >>> This is our first new project in over seven years. Abstract Wikipedia >>> was submitted as a project proposal by Denny Vrandečić in May of 2020 [2] >>> after years of preparation and research, leading to a detailed plan and >>> lively discussions in the Wikimedia communities. We know that the energy >>> and the creativity of the community often runs up against language >>> barriers, and information that is available in one language may not make it >>> to other language Wikipedias. Abstract Wikipedia intends to look and feel >>> like a Wikipedia, but build on the powerful, language-independent >>> conceptual models of Wikidata, with the goal of letting volunteers create >>> and maintain Wikipedia articles across our polyglot Wikimedia world. >>> >>> The project will allow volunteers to assemble the fundamentals of an >>> article using words and entities from Wikidata. Because Wikidata uses >>> conceptual models that are meant to be universal across languages, it >>> should be possible to use and extend these building blocks of knowledge to >>> create models for articles that also have universal value. Using code, >>> volunteers will be able to translate these abstract “articles” into their >>> own languages. If successful, this could eventually allow everyone to read >>> about any topic in Wikidata in their own language. >>> >>> As you can imagine, this work will require a lot of software >>> development, and a lot of cooperation among Wikimedians. In order to make >>> this effort possible, Denny will join the Foundation as a staff member in >>> July and lead this initiative. You may know Denny as the creator of >>> Wikidata, a long-time community member, a former staff member at Wikimedia >>> Deutschland, and a former Trustee at the Wikimedia Foundation[3]. We are >>> very excited that Denny will bring his skills and expertise to work on this >>> project alongside the Foundation’s product, technology, and community >>> liaison teams. >>> >>> It is important to acknowledge that this is an experimental project and >>> that every Wikipedia community has different needs. This project may offer >>> some communities great advantages. Other communities may engage less. Every >>> language Wikipedia community will be free to choose and moderate whether or >>> how they would use content from this project. >>> >>> We are excited that this new wiki-project has the possibility to advance >>> knowledge equity through increased access to knowledge. It also invites us >>> to consider and engage with critical questions about how and by whom >>> knowledge is constructed. We look forward to working in cooperation with >>> the communities to think through these important questions. >>> >>> There is much to do as we begin designing a plan for Abstract Wikipedia >>> in close collaboration with our communities. I encourage you to get >>> involved by going to the project page and joining the new mailing list[4]. >>> We recognize that Abstract Wikipedia is ambitious, but we also recognize >>> its potential. We invite you all to join us on a new, unexplored path. >>> >>> Yours, >>> Katherine Maher >>> >>> Executive Director, >>> Wikimedia Foundation >>> >>> [1] https://meta.wikimedia.org/wiki/Special:MyLanguage/Abstract >>> Wikipedia/June 2020 announcement >>> [2] >>> https://meta.wikimedia.org/wiki/Special:MyLanguage/Abstract_Wikipedia >>> [3] https://meta.wikimedia.org/wiki/User:Denny >>> [4] https://lists.wikimedia.org/mailman/listinfo/abstract-wikipedia >>> -- >>> >>> Katherine Maher (she/her) >>> >>> Executive Director >>> >>> Wikimedia Foundation <https://wikimediafoundation.org/> >>> >>> _______________________________________________ >>> Please note: all replies sent to this mailing list will be immediately >>> directed to Wikimedia-l, the public mailing list of the Wikimedia >>> community. For more information about Wikimedia-l: >>> https://lists.wikimedia.org/mailman/listinfo/wikimedia-l >>> _______________________________________________ >>> WikimediaAnnounce-l mailing list >>> wikimediaannounc...@lists.wikimedia.org >>> https://lists.wikimedia.org/mailman/listinfo/wikimediaannounce-l >>> >> >> >> -- >> * I use this address for lists; send personal messages to phoebe.ayers >> <at> gmail.com * >> > -- * I use this address for lists; send personal messages to phoebe.ayers <at> gmail.com * _______________________________________________ Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l New messages to: Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, <mailto:wikimedia-l-requ...@lists.wikimedia.org?subject=unsubscribe>
