the fight against hate reporting system

37
The Fight against Hate Reporting System Dr Andre Oboler CEO, Online Hate Prevention Institute © Andre Oboler, 2015 12 th biennial Seminar on antisemitism

Upload: andre-oboler

Post on 17-Sep-2015

75 views

Category:

Documents


1 download

DESCRIPTION

A presentation by Dr Andre Oboler, CEO of the Online Hate Prevention Institute, to 12th biennial Seminar on antisemitism of Tel-Aviv Univeristy's Kantor Center for the Study of Contemporary European Jewry.This presentation presents the Fight Against Hate Expert System and what becomes possible with transparency into the issue of online hate.

TRANSCRIPT

  • The Fight against Hate Reporting System

    Dr Andre Oboler

    CEO, Online Hate Prevention Institute

    Andre Oboler, 2015

    12th biennial Seminar on antisemitism

  • Lets change our perspective Our work, even in this seminar, has focused on

    individuals and specific examples of antisemitic propaganda

    This work is critically important, but it is not sufficient. We also need a big picture view.

  • An invitation to the balcony This presentation contains real and interesting data

    about online antisemitism. This data is not the really important thing in this talk.

    Step back, and the implications of the tool presented here, from which the data is gathered, will emerge

    A revolutionary change to our efforts in combating antisemitism is possible

    Join me in looking at the big picture, the possibility of systemic change. Lets talk about it, and see about making it happen.

  • WHAT WOULD WE CHANGE?

    Data & Analysis to plan mitigation & prevention

  • For off-line incidents, data is available. Based on this important data, action to mitigate and tackle antisemitism can be planned. There is no online equivalent.

    Need: Data & Analysis to plan mitigation & prevention

  • At the Global Forum to Combat Antisemitism we will release a report based on data From our new FightAgainstHate.com reporting tool. Here are some the results:

    Need: Data & Analysis to plan mitigation & prevention

    23%

    41%

    36%

    Antisemitism by social media platform

    Facebook

    YouTube

    Twitter

    5%

    12%

    49%

    34%

    Antisemitism by classification sub-types

    Promoting violence against Jews

    Holocaust denial

    Traditional antisemitism (not Israel-related)

    New antisemitism (Israel-related)

    Note that new antisemitism is significant, but much less than traditional antisemitism

    Sample size: 2024 items

  • Drilling deeper the results are even more startling. We see that different kinds of Antisemitism are more prevalent on different platforms. Prevalence is a combination of what users upload, and what action the platform is taking to remove such content.

    Need: Data & Analysis to plan mitigation & prevention

    16

    27 72

    Promoting violence against Jews

    Facebook

    YouTube

    Twitter

    42

    105

    44

    Holocaust denial

    Facebook

    YouTube

    Twitter

    214

    253

    120

    New antisemitism

    Facebook

    YouTube

    Twitter

    137

    433

    167

    Traditional antisemitism

    Facebook

    YouTube

    Twitter

  • What does this tell us? The view on the dance floor:

    We see specific areas such as violence on Twitter, or Holocaust denial on YouTube which need urgent attention.

    We see a more even distribution in New Antisemitism, perhaps indicating that all platforms are reluctant to take this for of antisemitism down, and this is the real split of where the antisemites are active.

    We can hypothesis that Facebook is doing a better job than YouTube and Twitter at removing hate (except new antisemitism)

  • What does this tell us? The view from the balcony:

    Being able to make such graphs and comparisons is new now we have transparency on social media antisemitism

    We have the list of 2024 items that make up the sample Allowing analysis of messages, sources, networks etc

    We can monitor if / when they are removed Quality control data on platforms responses

    The data comes from public reporting, so we can learn what content is causing the most distress (leading to the most reporting) overall, or by country Content that most urgently needs to be tackled / campaigned on

    The data is always live in the system, and we are offering supporting experts such as NGOs, academic researchers, community organisations, government agencies, access to it This can empower many approaches and actions to tackle online

    antisemitism

  • WHAT WOULD WE CHANGE?

    Empowerment for Civic Action

  • Meaningful action by civil soc. & individuals I do believe and I have expressed that in the Parliament,

    that Neo Nazis, antisemites, racists and intolerants are using internet and new technologies, in a very effective way, spreading the virus of hate. The issue is how to react and how the civil society or the individuals can make their ideas a reality in order to tackle this issue. We need democratic action especially in terms of social media and internet. - Kostas Triantafyllos, former Member of Hellenic Parliament & rapporteur of the Anti-racist bill of Greece (e-mail, 9/5/2015)

    Need: Empowerment for Civic Action

  • The current situation Users feel ignored and frustrated

    1. User report content 2. The report is typically rejected 3. User gets frustrated and doesnt waste their time reporting next time

    they see something

    Organisations, particularly Jewish ones, want to engage with

    tackling online antisemitism, but dont know how.

    New grassroots online initiatives exist doing mass reporting. These are isolated from each other and dont feed into our systemic view of the problem.

    Grassroots efforts only have success with things the platforms agree should come down. They cant convince the platform to recognise new forms of antisemitism, or that specific old forms are being wrongly rejected.

    Need: Empowerment for Civic Action

  • Empowering Individual and Civil Society Action An alternative place to report,

    where items are logged, and a lack of action by the platform will be seen, motivates reporting

    Online groups can contribute to a global effort, confident that the hard cases they cant progress will be tackled by others

    Communities / civil society organisations can use the tool as a closed community for their members as the technical part of an online campaign

    Empowerment for Civic Action

  • Resource booklet to support campaigns

    Empowerment for Civic Action

  • WHAT WOULD WE CHANGE?

    Empowering Government Action

  • Governments need help

    Empowering Government Action

    Governments are increasingly demanding more accountability from platform providers for the content spread on their platforms. The argument that it is all the users responsibility is dead

    Some governments are introducing systems of fines These however need to be fair. Ideally it should be like pollution,

    where the fine is for dropping below a known standard, and the fact this occurred can be empirically demonstrated.

    Some courts have imposed penalties when platform providers have not done enough to address problems Courts can help define the standards through case law, but they

    need on which it do it e.g. How many cases are taking for than 3 months to resolve?

  • WHAT WOULD WE CHANGE?

    Focus on Hard Cases

  • The hard case problem Monitoring by experts is not enough. There is too much data

    YouTube: 2,056,320 videos are uploaded each day Facebook: 350,000,000 images are uploaded each day

    Some cases are obvious The public can identify them and report them Simple training will allow junior staff at social media companies to respond for these items there is no problem, except monitoring the speed of the response

    Other cases are less obvious The public may not be able to explain why the content feels antisemitic Platform staff may not have enough expertise and may reject the reports Experts need to know the case exists so they can explain it to the platforms

    The challenge is separating easy and hard cases and using limited resources wisely

    Focus on Hard Cases

  • Implication of Hard Cases

    Platforms should initially have more time for dealing with hard cases... i.e. new forms of antisemitism, but after it is explained, next time there should be a different expectation on very similar content.

    The failure to learn in a systematic way should be penalised.

    We need to track what the platforms should have learned at the global level

    Focus on Hard Cases

  • Dealing with the hard case problem One way to filter for the hard cases is to look at the items that

    arent coming down, despite many reports by people who in other cases tend to get it right i.e. Their other reports are accepted by the platforms

    We need mass participation both for reporting and for reviewing the reports of others (crowd sourcing) for this to work

    We need participation in each country (even if the same content is being reported) in order to bring the case into as many jurisdictions as possible, enabling more officials and more experts to tackle it

    We need expert to go through the hard cases and within the framework of their existing organisations, tackle some of them

    Focus on Hard Cases

  • A look at FightAgainstHate.com

  • Overview

    A Crowd Sourcing solution Overcomes difficulties around recognition, with enough eye

    balls, someone will recognise the problem

    Creates problems of bias, ignorance and game playing

    An artificial artificial intelligence (AAI) solution Effectively using humans as agents, as we would do with

    software agents

    We have multiple people assessing content, but maintain a model of trust in those doing the assessing

    This provides quality control and prevents game playing

    A look at FightAgainstHate.com

  • How it works Users report the URL of online hate

    They classify it

    They review items reported by others

    In both review and classification they express an opinion over the nature of the content, and their confidence in their opinion

    A points system favours strong opinion that are correct, and weaker opinions that are wrong

    Points are adjusted on verification of the content by the system (based on the crowds opinion) or by an expert

    A look at FightAgainstHate.com

  • THE EXPERT SYSTEM...

  • Examples of Statistics Available

    Items reported over time

    Increase in items by social media platform

    Trends over time by hate type

    Time taken to remove items

    Average time to remove item by platform

    Trends in reports by state/country

    Responsiveness by platform

    We want to work with supporting organisations to see what information would be useful.

  • 5%

    12%

    49%

    34%

    Antisemitism by classification sub-types

    Promoting violence against Jews

    Holocaust denial

    Traditional antisemitism (not Israel-related)

    New antisemitism (Israel-related)

    Imagine graphs like this, by country, generated live

  • 16

    27

    72

    Promoting violence against Jews (breakdown of items by social media platforms)

    Facebook

    YouTube

    Twitter

    Or a graph like this to start a conversation with Twitter about needed improvements

    10 6

    Facebook

    pages & groups

    users & items

    1

    26

    YouTube

    users & channels

    videos

    19

    53

    Twitter

    users

    tweets

  • An offer... Using the system to run campaigns to collect data, and seeing

    the raw data your campaign collects, is free talk to me about running a campaign

    We are also looking for supporting organisations, we have 26 in Australia but would like to take that global Supporting organisations will have access to the expert system, and

    will provide feedback on it, and a statement of support for fightagainsthate.com

    Access to the expert system will be free for supporting organisations until October, from October fees will apply to some expert system features, these will help cover the costs of running & improving the system

    We are also looking for donations to improve the expert system; to make the system multi-lingual; and more. Any leads would of course be appreciated.

  • Thank you.

    Questions?

    CONTACT DETAILS

    Dr Andre Oboler

    CEO, Online Hate Prevention Institute

    ohpi.org.au & fightagainsthate.com

    Twitter: @onlinehate

    www.facebook.com/onlinehate