Social Network Analysis of Disinformation/Influence Operations and Bots Cross-Platforms on Amber Heard.
Twitter, Reddit, YouTube, Instagram, Change.org, Facebook, TikTok, Tumblr for 8 Social Media Platforms.
We're Worldwide Independent Researchers with over 20 data analysts, scientists, and observers analyzing this data for over a year. Research and development is ongoing, and this public operation spanning for years continues. We invite researchers, an ethical public, and students to study this public case example.
Data collected on Amber Heard is primarily related to the years of the operation of 2018-2022, but is fully collected from 2008-2022.
This is the most detailed public analysis of a disinformation operation and the most voluminous yet - showing the impact of Category 6 influence, globally affecting the public, businesses, nonprofits, reporters, witnesses, social connections, well-being, and entities beyond the victim and human rights activist Amber Heard.
This dynamic project is in-progress, prepared for review, and ready to be built upon.
Our Study creates Precedents and Foundations to help victims of cyberabuse, bots, domestic abuse, coercive control, crime, hostile environments, and Disinformation Operations to Improve Lives.
We want to save lives and help partners create systems to help online - e.g., specialized and accurate rescue, quality custom studies, data analysis, Social Network Analysis, forensics, research, situation awareness, data science projects, prevention, and rescue/public safety technologies - with focus on the victim primarily and her environment.
- Public Research:
This study and its data is open for saving lives, innovation, scientific advancement, developing solutions, maximizing knowledge, blueprints, and archiving for long-term benefit.
- If a file of analysis is too large to preview on github, please download, or preview in this google folder.
- Full Data sharing is in-progress on Kaggle, Google, Archives, Scribd and public research data repositories.
Amounts of Data Collections [Updates]:
- Twitter: 2008-2022 Data on Amber Heard.
1,732,916 Tweets and 459,228 Accounts Jan 2018-April 2022.
985,400 Tweets and 275,567 Accounts Jan 2018-April 2021. 497K Tweets with 182K accounts Jan-April 2022, (2) Over 880,000 Tweets of Top Users, (3) 700K+ Retweets, Links, (4) 56K Liking Accounts 2021, (5) 104K Quote Retweets, 57K Accounts of AH Timeline 2017-2022, (6) 134K Comments on Amber Heard Profile 2017-2022, (7) 2008-2017 with 698K Tweets on Amber Heard, Cluster Analysis 2008-2022 - Reddit: 164,530 Contributions, 15,896 Submissions, 71,319 Accounts, Links - 2018-2021
- Instagram: 1,751,113 Comments, 193,967 Posts, 717,311 Accounts (Posts made by 36,137 accounts and 681,174 commented), Links - 2018-2021
- YouTube: (1) 2,176,926 Comments (1,693,341 Comments with 483,585 Replies), 6,893 Videos (with dislike ratios), 1,152,868 Accounts, Cross-Platform Links, (2) 11,152 videos, 324,029 comments (2 collections)
- Change.org: 28,952 Comments, 181 Petitions, 2,117 Milestones, Cross-Platform Links
- Facebook: Page Comments and Reviews, Links in Groups for Cross-Platforms
- Tumblr: 35,898 Tumblr Posts Data with "Amber Heard" tag 2009-2022
- TikTok: Data of daily feed videos under 57 related hashtags to AH
- Crossplatform for Groups in Facebook, Reddit, and Channels of YouTube
Use for Timeline Correlations to Risks, Preliminary Effects, and Origins
- Analysis of Social Media Platforms - Complex Study with Detailed Python Notebooks
Twitter, Reddit, Instagram, Change.org, YouTube, Facebook, Cross-Platforms
Peaks, Anomalies, Timelines, Patterns, Statistical Models
Timings, Graphs, Bot Networks, NLP/NLU and Sentiment Analysis
Banned, New, Unverified Account Layers
Investigations of Peaks, Accounts, Comments, Postings
Repeated Texts with Timings, Multiple Accounts, Community Detection
Coordination and Gamifications
Threat Analysis and Negative Texts Heat Maps
Metrics, Likes/Dislikes, UpScores/DownScores, BotScores
Links Analysis, Urls, and NLP Analysis Cross-Platforms
Clusters Analysis, Relational Dataframes, Wordclouds, OSINT - Anomalies and Comparisons
- Reports on Analysis and Platforms:
Summaries of Analysis, Years, Timelines, Accounts, Timings, NLP, Peaks
Showing Coordinated Activity and Anomalies of Disinformation Operations - Images/Highlights/Figures
- Data Samples
- Banned, Deleted, Suspended Accounts
- Research Papers
- NLU Classifications, Monitoring, AI
- Positive NLG Compliments, Prompts, AI
- Analysis Guides
- Background
Compliments Generation on Amber Heard based on Jason Momoa
https://github.com/semiosis/prompts/blob/master/prompts/compliment-generation-based-on-a-celebrity-1.prompt
- 10,000 Compliments on Amber Heard made in 2 hours: https://asciinema.org/a/w5ayuAEGGpgsxTCKWbAGRPfvZ
This example shows positive articulate language towards Amber Heard using artificial intelligence language models and shows helpful, supportive relationships. - Funny 3-Way Bots chat with Actress Amber Heard, Actor Tom Cruise, and Director Christopher Nolan demonstrated at: https://semiosis.github.io/apostrophe
- Code, Demos, and Videos from AI Researcher at: https://github.com/mullikine/positive-nlg-compliments
- Semiosis is a free and open-source curation of prompts for OpenAI's GPT-3/Codex, EleutherAI's GPT-j, AlephAlpha's World Model and other language models.
If you're a humanitarian, researcher, social enterprise, organization, student, or academic helping female victims of abuse, coercive control, domestic abuse, crime, bots, operations, violence, harassment, or hostile environments - we welcome collaboration with you.
Our studies and data can be used for Social Network Analysis, Data Analysis, machine learning, Natural Language Processing, artificial intelligence, Data Science labeling, anti-abuse, rescue projects, anti-crime, timelines, OSINT, simulations, and monitoring systems to help other victims and their lives.
Since this data is high volume about domestic abuse and gender, with an actress/model and film industry element, it's highly relevant for classification and data science projects on those topics.
Please contact us for collaboration or if with questions on obtaining more data and code.
Although it's abuse of process, coercive control is becoming illegal in more countries and states/provinces worldwide. Harrassment is illegal.
To this degree, severity, and volume of intimidation, witness tampering and extortion is already illegal.
- If you're a victim needing help against cyberabuse especially in online domestic abuse, operations, crime, and hostile environments, please use this example for your own awareness and contact us.
All collaborators, researchers, and observers are required to sign, acknowledge, agree to "Do No Harm" agreements. If you are one of the victims, businesses, nonprofits, or entities listed in this study and want to collaborate with us, please contact us.
We make this study and data fully open to Public Research. We do not include the full code, however, we make data open on data repositories to protect the individual in this study, to make a long-lasting impact for learning in the scientific world, and to prevent harmful operations. As a precedent, other researchers and students should use this GitHub and study as a basis for their own social network analysis projects.
We draw attention to this project months before in-person appearances for safety.
On GitHub:
- Metrics of Twitter Data and Clustering Data 2008 - April 2022
- Search queries on YouTube, video detail lists
- 5 Months of Data for Cross-Platform Analysis:
- 307,649 Tweets of Twitter, Texts of Accounts in Coordination, 674,469 Retweets, and All Listed Accounts
- 43,509 Contributions of Reddit and All Listed Accounts
- 394,237 Comments of Instagram and All Listed Accounts
- 28,952 Comments of Change.org, 2,117 Milestones, and 181 Petitions Data
- 6,893 YouTube Videos Data, 424,457 Comments of YouTube, "Adapt & Survive" 130K Comments, All Listed Accounts
- 5.9K+ Labeled NLU Data of Support, Defense, Compliments and Training Data
- Threat Analysis Data and NLU Test Data for Instagram, Reddit, and Twitter
- Anomolies Data Samples
- 10,168 Compliments of Amber Heard from AI NLG Prompts
- Links/URLs Lists of comments/postings of Instagram, YouTube, Twitter, Reddit, Change.org
This Independent Study is open for Public Research, scientific learning, and archiving.
Full Data is on Kaggle, Google, Archives, and research data repositories.
- "Research data can save lives, help develop solutions and maximise our knowledge. Promoting collaboration and cooperation among a global research community is the first step to reduce the burden of wasted research."
- "When researchers make their data public, they increase transparency and trust in their work, they enable others to reproduce and validate their findings, and ultimately, contribute to the pace of scientific discovery by allowing others to reuse and build on top of their data."
Across-Platforms Analysis:
Analysis is in-progress across-platforms, going beyond any previous public studies.
- Data of 5 months of December 2020 to April 2021 is provided on GitHub for open peer-review and research studies. Full Data for more independent cross-platform analysis is provided in research data repositories.
- Note: Peaks, Heat Maps of Threats, coordinated bot activities, repeated same texts, and more anomalies show for 2018-2022 continuing, especially February 2020 onward as seen in analysis files.
Not all analysis is given for researcher privacy for published research papers and review of the operations.
The Academic API was used for Retweets and InformationTracer for CrowdTangle API.
- We show increases of threats and escalations - even in spite of her winning in cases or gaps of announcements.
- Threat analysis and our studies shows the timelines to escalations of risks with keywords, phrases use. Hundreds of thousands of threatening texts exist against Amber Heard on 6 platforms. In social dynamics, within automated simulations caused by services are percentages of incited people to harm. In examples, within tens of thousands of repeated same texts on YouTube (e.g., "Adapt and Survive"), false interactions on all platforms, and cryptic bot networks are people brought into them to threaten Amber Heard.
- Experts saw cyber crime and suspicious activities needing flagging for removal by security. See more in Study Case.
- Clustering analysis with cyber intelligence research reveals scientifically operations to cause distress beyond searching manually. Clustering further shows sports bots networks, movie networks, and 3,658 other clusters 2018-2022 that partly contain the Aquaman petition, and astroturfing e.g., Reddit NLP sandbox.
- Multiple layers of companies, services, are likely. Along with the case's circumstances of intimidation, violence, missing people, witness tampering, high wealth of opposition, Hollywood Fixers, and suspicious activities, the online world similarly has been made threatening to exploit the Data Void.
"Data voids are not unique to search engines; they occur on social media platforms, too."
Origins, Risks, Situation, and Private Sector Operations:
- After reviewing analysis and data quantifications, explanations of origins and motive can be discovered.
Folder 'Study Case - Effects, Risks, Origins' shows dangerous Private Sector operations. - 'Subjective Preliminary Claims' sources explanations to layers of services, companies, decades of fixers, coercive control, social media building, insider threats, and malice where tampering and extortion using cyber activities can come from.
- Amber Heard's August 2020 Counterclaims on Bots focused on Correlating Edited Media Leaks and Top Posting Twitter Accounts Layer only: https://www.scribd.com/document/473092071/cl-2019-2911-def-counterclaims-8-10-2020
- Multiple Research Papers on Disinformation/Influence Operations Cross-Platforms
- Timelines of Situations of distress to Data and Studying Technologies (e.g., data analysis, data science, AI/ML, NLU/NLG, content review systems).
- Influence Operation of 6,000 Twitter accounts found by BotSentinel: https://youtu.be/bb2bC04OEPw
"Disinformation in the Private Sector: The Price of Influence"
There have been victims of disinformation, intimidation, witness tampering, crime, and influence operations in the Private Sector for decades. Operations can mirror coercive control strategies of multiple types of abuse. This open-source study can be used to help others by creating awareness, warning, and guides without needing to start from scratch.
- Hired Services - New Accounts Layers, Aged Accounts, Created News-Sites:
https://go.recordedfuture.com/hubfs/reports/cta-2019-0930.pdf - Psy-Group's Operations and Bots - https://www.newyorker.com/magazine/2019/02/18/private-mossad-for-hire
- Devumi Services and Bots for Online Influence Industry and Public Figures - (1) https://www.thebureauinvestigates.com/stories/2017-12-07/twitterbots
(2) devumi-a-firm-that-sells-fake-twitter-followers-to-be-probed-by-new-york-attorney-general - BlackCube vs Actresses - https://www.newyorker.com/news/news-desk/harvey-weinsteins-army-of-spies
- Medium "Bots Created Anti-Amber Heard" Operations - https://medium.com/@aquaman-bots/how-social-bots-created-an-anti-amber-heard-aquaman-campaign-e68e16637d3a
- Video - Rose McGowan vs BlackCube - https://www.youtube.com/watch?v=tv6ewJQ7L3k
- Actress Meghan Markle Operations - (1) https://laptrinhx.com/meghan-markle-s-twitter-bot-network-the-whole-thing-is-a-bit-insane-2560404981 (2) https://www.newsweek.com/meghan-markle-troll-accused-posing-black-woman-second-account-suspended-twitter-bot-sentinel-1652271
- Jessica Reisinger-Raubenolt & Lilia Operations - https://www.tampabay.com/news/crime/2021/08/05/cameron-herrin-went-to-prison-for-a-tampa-crash-were-the-tweets-that-followed-real
- Amber Heard, Cyrillic Spiders from Mars - https://www.thegeekbuzz.com/the-basement/cyrillic-russian-spiders-from-mars
Worldwide Data Scientists, Researchers in Cyber Intelligence, PhDs, Data Analysts, Artificial Intelligence Researchers, and a former Google Research Intern collected and analyzed data. We use the scientific method of testing.
- Category 6: The Breakout Scale: Measuring the Impact of Influence Operations by Ben Nimmo in Social Network Analysis Standards:
"Category One operations only spread within one community on one platform, while Category Two operations either spread in one community across multiple platforms, or spread across multiple communities on one platform. Category Three operations spread across multiple social media platforms and reach multiple communities.
Category Four operations break out from social media completely and are amplified by mainstream media, while Category Five operations are amplified by high-profile individuals such as celebrities and political candidates.
An IO reaches Category Six if it triggers a policy response or some other form of concrete action, or if it includes a call for violence." - Scientific Method: Observation, Question, Hypothesis, Experiment, Results, Conclusion
- Limit Variables and Repeat Same Methods on all platforms.
- Verifiable IDs of postings and accounts, APIs, multiple forms of documentation.
- Cross-checking multiple times with full field JSONs and CSVs with processes in data science. - Situation Awareness:
Objective Tactical (short-term): situational assessment - situation awareness
Strategic (long-term): sensemaking - understanding
Scientific (longer-term): analysis - prediction
We welcome Researchers to work with this Data on Amber Heard to create Blueprints on how Bots and Disinformation/Influence Operations work.
- Continuation: Research and building technologies is Ongoing as the operation is Continuing over years and this is a Public Github Example.
We are continuing analysis and data collections, including creating technologies for systems on coercive control.
E.g., simulations/monte carlos with multi-agent modeling of patterns of coercive control from victims' experiences shown in media - examples are Hannah Clarke and others. Data analysis connecting timelines of documentation is encouraged for new systems.
“Since the fate of the world is non-existence, Since You Exist, Be Merry.” - Persian Mathematician Omar Khayyam - explained by Amber Heard.
If interested in citing our research project in your studies, publication, journal, database, media outlet, or article, please provide a link to this GitHub repository with its title and contact us for collaboration.
- Please ask to collaborate with us, ask us questions, or provide suggestions on our Q&A, [email protected], LinkedIn, or contacting us individually.
Researchers ethically did this study independently with hopes to save a life and to create knowledge that can help others who need protections from operations, abuse, violence, and bots.
We encourage you to build upon our studies and analysis for your own projects.