HRPC @ IETF105 Agenda HRPC IETF 105 # Welcome and introduction (5 minutes) Scribe, Note takers Agenda Review Research Group status # Talk by JC de Martin (30 minutes) Ethical and Socially Aware Data Labels: https://nexa.polito.it/node/1515 # Updates/discussions on research group drafts (30 minutes) draft-irtf-hrpc-association, Joe Hall and Stéphane Couture draft-irtf-hrpc-political, Niels ten Oever (remote) # Update/discussion on draft not yet adopted by the RG (15 minutes) draft-guerra-feminism, Juliana Guerra and Mallory Knodel # Update on the Guidelines draft and short presentations from reviewers (20 minutes) draft-irtf-hrpc-guidelines, Gurshabad Grover and Niels ten Oever (remote) # AOB (10 minutes) --- hrpc.io Presentation: Ethical and Socially Aware Data Labels: https://nexa.polito.it/node/1515 (JC de Martin, Nexa Center for Internet & Society) * Slides https://datatracker.ietf.org/meeting/105/materials/slides-105-hrpc-demartin-00 * Authors: Elena Beretta, Antonio Vetrò, Bruno Lepri, JC De Martin * crucial role of data in digital systems * Cathy O'Neal, WMD: if admission models trained on American univs before 1960s, very few women * to avoid unintended negative effects: care is needed at all stages of design & dev * steps * data collection * data usage * label datasets to help computer scientists * certain data has characteristics that may be dangerous * risk of discrimination e.g. uneven distribution in gender balance in input data * useful for engineers and third parties (certificator) * other initiatives: * dataset nutrition label project * datasheets for datasets * Three building blocks for EASAL * disproportionate datasets * correlations and collinearity * data quality * ISO/IEC 25012 and 25024. Quality dimensions: accuracy, completeness, consistency... * Testing the EASAL approach on a real case * case: credit card default dataset * taiwan, april 2005 - sept 2005 * age, credit data, demographic attributes (not race but gender etc to discriminate) * Disproportionate datasets * Histogram of variable age (graph) Density to Age * not necessarily bad, just something to keep in mind * 60% of this dataset are women, 50% single vs married, 46.7% have attended college. Can have implications * Correlation and collinearity * correlation between payment default and education, gender, marital status * Data Quality * accuracy and completeness - quite good * EASAL approach could help datasets to be more aware of the potential problems BEFORE using the dataset, to be made more aware of the risks of downstream unintended problems. * QUESTIONS? * Jonathan Hoyland: is labelling data not going to lead more discrimination? * Emphasis on the word "unintended". * follow up: could adding noise help? To break the collinearity? * Joe Hall: DARPA has a version of Android - Privacy Enhanced Android. Lets devs tag data for risk. Does this work relate to that? * good idea Draft Presentations: * Draft-association update https://datatracker.ietf.org/meeting/105/materials/slides-105-hrpc-demartin-00 * Joe Hall (CDT) and Stephane Couture (York U) * no update since last time * intro for SC * Background: * Niels and Gisela handed it off at IETF 104 * about ~75% done * Summary * research question: how does the architecture of the Internet enable/inhibit the right to freedom of assembly and association? * methodology: test the relationship between protocols and association, assembly through cases * 7 cases * Question 1: causality * difficult * 8280 hasn't done this * drop? just want to describe the relationship * Question 2: which cases? * conversing, p2p, identity * what else? gatekeepers like app stores? Apple can rm Twitter * social case? * Question 3: assocation on proprietary platforms * implicitly critical of large platforms such as social media companies * ... but also Twitter helps revolutions? * Proposal is to soften this critical stance to highlight that humans have adapted platforms to their association needs * freedom of disassociation? * Question 4: who is the audience for this draft? * Minor changes * needs to be less normative, more descriptive * Question/Comment 5: * great work, but careful about going down the DDoS rabbit hole especially as HRPC has had that conversation early on and it was not helpful. * Mallory echoes that point. * Joe: we want to use it as a discussion point. * Mallory: perhaps go back to the list. * CCS: you can find the discussion starting here: https://mailarchive.ietf.org/arch/msg/hrpc/Bs-AQftKY8b8hwXCjDxXDuudO3s * Comment, Juliana: * important to keep in mind the end user * Comment by Leandro * Need to have some governance across orgs. Case of instant messenger: not federated or interoperable * -03 in a month, questions to the list * Draft political: https://datatracker.ietf.org/meeting/105/materials/slides-105-hrpc-draft-irtf-hrpc-political-03-00 * by Niels * review by Julia Guerra and Michael Rogers * overview of changes * where is the document now? * Mallory: can you talk about the recent discussion on the list? * Andrew Sullivan said that there might be consistency problems * conversation might not have been about the most current version * Niels thinks it's ready * Joe Hall: want to give one last review * Gurshabad Grover: analysis of positions taken by IETF/IRTF in scope? * Niels: tech and politics and community positions. Tried to document positions observed in the community as they were. * not sure how to further analyze the positions * Avri: can people please do a thorough read through? * Joseph Potvin: don't see a reference to RFC 2119. Provides formal definitions of words. * Mallory: does this draft achieve capture of discussion on political boundaries of protocols? * Draft feminism by Juliana Guerra: https://datatracker.ietf.org/meeting/105/materials/slides-105-hrpc-draft-guerra-feminism-01slides-105-draft-guerra-feminism-01-00 * two different problems * user level * Internet threats considering gender * how are those threats be related to protocols? * in draft -01: Feminist Principle of Internet capture the idea that different social contexts may determine people's capacities to exercise rights * IETF level * RFCs and IDs * Intersectionality * oppressions can be structural and dynamic: example of the presenter's different privileges in different contexts * Matrix of domination * Access Principles: * how protocol related concepts ('end-to-end principle', 'interoperability', etc.) capture (or not capture) access * Matters related to IETF as an org removed from this version * Next steps: * participative process * presenting at at Citizen Lab Summer institute * Mallory: also participating in mailing lists about the draft * Tim Costello @ mic * Niels: how are you going to operationalize how power is understood? power == money in capitalism. * have to scope it down though it is hard with intersectional feminism * Jonathan: privacy? if you can't discriminate between people then you can't, well, discriminate between them. Then is privacy a more important problem? * Feminist principles. The issue is how can we make our own choices on the Internet? * Mallory: part of the point is to not have to hide that you're LGBT, or you're a woman. Positive expression of autonomy. * Sophie: how has this draft been received? feminist protocols or just impact on feminism of protocols? Also, what can be learned from feminist infra? * Feminist protocols is the dream! * Hernani Marques from PEP foundation: concrete cases of oppression. Connect with Sarah Jamie Lewis who has written a book on oppression (in Vancouver) * Stephane Couture: thank you for bringing this work. Alternative networks by Leandro * Guidelines for Human Rights Protocol and Architecture by Gurshabad Grover https://datatracker.ietf.org/meeting/105/materials/slides-105-hrpc-slides-for-update-on-draft-irtf-hrpc-guidelines- * Draft updated in May, thanks Joe * Addressed Avri's review * Next steps * Niels * Joe * Mallory: human rights review team mailing list officially shut down. Gurshabad Grover on human rights review: not much experience since last meeting.