- From: Mike Gifford <mike.gifford@civicactions.com>
- Date: Fri, 28 Apr 2023 09:42:04 -0400
- To: "public-civics@w3.org" <public-civics@w3.org>
- Message-ID: <CAL71Ph5Oi2G10-W_aEXXVOJmoV0MJheERFrTLdY-mn_PFx2DYA@mail.gmail.com>
Hey Adam, When it comes to gauging public opinion, my preference is to skip polling and move right to solutions like https://pol.is/home Taiwan has done some great work with this to help build on similarity rather than difference: https://blog.pol.is/pol-is-in-taiwan-da7570d372b5 There are likely ways to gain additional insights by applying ML to the content produced by citizens. Being able to move beyond sentiment analysis to actually group and summarize free-form expressions of interest is certainly worth exploring. It may well allow for some more qualitative feedback, which is often missing from more numerical poling approaches. I still wonder whether adding the BS & hallucinations from ML is going to get us further away from Benjamin Disraeli's "Lies, Damn Lies and Statistics". Will it reinforce the bias that it has learned by scanning the last 100 years of human text? There are risks in all of this, and not all algorithms or datasets are built to avoid bias. Mike On Thu, Apr 27, 2023 at 10:09 PM Adam Sobieski <adamsobieski@hotmail.com> wrote: > Civic Technology Community Group, > > > > Hello. I recently received some feedback on these ideas which underscored > the importance of securing opinion-polling systems. > > Firstly, virtual opinion pollsters could be those parties which initiate > opinion polling processes, using telephone to call selected respondents, > sending emails to selected respondents, and so forth, as opposed to > traditional Web-based opinion polling where respondents initiate opinion > polling processes (including in ways which might not be impervious to > malicious actors or their bots). Anonymity in these types of systems, > where virtual opinion pollsters initiate the processes, would, seemingly, > require trusting the opinion-polling organizations using the AI tools. > Secondly, there exist solutions such as *Vocdoni* [3][4]. Iām looking, > once more, at these types of technologies ā which I think are designed > for multiple-choice polls and voting ā and considering whether or not > they might be able to support more kinds of items including, but not > limited to, open-ended natural-language questions. > > > > > > Best regards, > > Adam Sobieski > > > [3] https://vocdoni.io/ > > [4] https://blog.aragon.org/introducing-vocdoni-anonymous-voting/ > > > ------------------------------ > *From:* Adam Sobieski <adamsobieski@hotmail.com> > *Sent:* Tuesday, April 25, 2023 3:59 AM > *To:* public-civics@w3.org <public-civics@w3.org> > *Subject:* Artificial Intelligence and Opinion Polling > > Civic Technology Community Group, > > Hello. I would like to share some more artificial intelligence and civic > technology ideas with the group. > > Artificial intelligence systems, virtual opinion pollsters, can perform > structured, semi-structured, and unstructured surveys, questionnaires, and > interviews across a number of communication channels (e.g., Web-based > chatbots, email, telephone, Microsoft Teams, Skype, Facebook, Slack, Kik, > Telegram, Line, GroupMe, Twilio, WebEx, WhatsApp, Zoom, RingCentral, etc.). > > Recent advancements to artificial intelligence and natural-language > processing, e.g., text embeddings, are interesting to consider with respect > to the advancement of opinion polling technologies. With natural-language > processing, virtual opinion pollsters can perform open-ended questions [1], > e.g., follow-up questions which might explore rationales, justifications, > and argumentation of respondents' previous answers. > > In addition to being able to perform predefined lists, or sequences, of > questions, virtual opinion pollsters can traverse larger trees or graphs of > questions, with paths branching, or varying, based upon respondents' > answers. > > Thank you. Any thoughts on these topics? > > > Best regards, > Adam Sobieski > http://www.phoster.com > > [1] > https://news.gallup.com/opinion/methodology/406922/natural-language-processing-aids-open-ended-questions.aspx > [2] > https://news.gallup.com/opinion/methodology/233291/why-phone-web-survey-results-aren.aspx > > -- Mike Gifford, Senior Strategist, CivicActions Drupal Core Accessibility Maintainer https://civicactions.com | https://accessibility.civicactions.com http://twitter.com/mgifford | http://linkedin.com/in/mgifford
Received on Friday, 28 April 2023 13:42:22 UTC