Relevant Reporting

A controlled approach to content discovery is key to developing insights, sources and search techniques that allow the individual or a team to monitor a story, location or topic over time. This approach is replicable but would need to be tailored to the topic.

Understanding the Media Environment

You must first build an understanding of the media used by perpetrators, by their target audience, and by actors seeking to counter perpetrator messages, as well as to understand, where possible, the motivations of those involved in the dissemination of such content.

Case Study 1 below, for example, involves content shared by perpetrators as well as by activist groups seeking to raise awareness of the issue. Tracking the activities of both parties and understanding their motivations is central to building an understanding of the situation.

The key question to ask is where would someone involved in the creation of activist content in a certain location be most likely to share that content in order to meet their aims?

What motivates a perpetrator to publish publicly on social media, as opposed to privately on a content-hosting or archiving site? How would media attention or the policies of social media platforms themselves impact a publication strategy for a militant group, for example?

Over time an investigator or journalist can build an understanding of where, how and why content is posted. This understanding acts as a filter, allowing faster access to content that is closer to the original source.

This understanding must adapt to changing patterns, stricter social media monitoring by the platforms, or other changes.  

The approach to searching for footage from violent actors in conflict zones will often differ from that used to search for extremist content in Western nations. Searches may include a mix of traditional and fringe social networks, including:

• Twitter
• YouTube
• Facebook
• Instagram
• Snapchat
• Whatsapp
• Telegram
• Gab
• 4chan
• 8chan
• Endchan
• Reddit
• Voat

Different strategies will be needed for different platforms, based on the nature of the content, the violence, or the criminal activity being considered.

You must understand and adapt to online conversations to source relevant, original content quickly.

Ethics of Interaction

In order to track perpetrator conversation and publication activity, it is necessary to have active accounts on all relevant platforms and to build alerting methodologies using all available tools. From this monitoring process, hashtags, keywords, colloquialisms, and phrases can be harvested to allow for content searches.

While it is best practice to always, ordinarily, communicate your intentions to sources and uploaders, stating to them clearly the terms of use and the permissions being sought in the uploader’s language, the same cannot be said when it comes to perpetrator content.

Facebook community page administrators, activists groups and independent activists will be open to engaging with requests. However, reaching out to fighting groups and armed militias is discouraged. There is a risk the outreach may draw unnecessary attention to the researchers or to local sources that provided content or information as part of an investigation, or draw attention to victims visible in perpetrator content?

Perpetrator content is often publicly available, that is, published by a party, actor, or group for public dissemination, and often with the intention to do harm or for propaganda purposes. Observers should consider whether the benefits of publication or dissemination, whether by journalists or by human rights investigators, outweigh the risk that doing so will further the aims of the perpetrator.

If publication or dissemination is warranted, other considerations arise. Can you identify the victims to seek their consent before sharing the content? If you can identify them, could others, potentially to their detriment? Would publication or dissemination of the content put the victims, their families, or their wider communities at risk? Further discussion on this topic can be found in the ethical use of perpetrator content section.

Consider also the risks to which you may be exposing a source in reaching out to them publicly, or even privately. Set out rules of engagement for such public outreach for your team.

Also set out ethics guidance on how your organization’s team members will behave and be identified on social platforms, to avoid them misrepresenting their role or intentions online. It is important to be transparent and upfront in conversations with sources or content uploaders.

On traditional social platforms (Twitter, Facebook, etc) create user accounts that clearly identify who is behind the account and explain why you are reaching out to someone. Using closed platforms and messaging services, such as Telegram, WhatsApp, or similar platforms, is different: the user accounts are more easily made anonymous. When trying to gain access to groups or chats on such apps, does each member of that group understand the implications of sharing content with you?

Where publicly contacting a source may put a journalist or investigator at risk, shared accounts under the organization’s name may be appropriate.

If, while reaching out, you don’t identify either your name or your organization’s name, the work effectively becomes an undercover investigation. Extraordinary circumstances are required  before such an approach is taken and, again, your organization should set out clear rules for the team to avoid the possibility of deceiving a source or misrepresenting the work.

Role of Citizen Video

Content discovery

The outcome of an analysis of the media environment should be:

• An understanding of the platforms used
• An understanding of the language used
• An account of relevant:
  • Hashtags
  • Keywords
  • Colloquialisms
  • Phrases
• A set of accounts to follow and monitor

Below, we will set out the methodology for sourcing, archiving, and making sense of content on these accounts.

Developing sources

Establish reliable sources to more quickly find content, daily leads, and people who can provide corroborating information to help in the verification.

• Reliability cannot be assumed. It must be established over time. Reliable accounts lead to other reliable accounts. Factors to consider include:
• Can content from the source be independently verified?  
• Does a source’s reporting tally with multiple independent sources?
• Can the source be contacted directly?
• How do they respond to being challenged on a claim?
• Are they forthcoming with information that can be independently verified?
• Are they transparent about their motivations?

Wherever possible, contact the source to learn where their information and content comes from. Is it realistic that an individual could provide all of the content on a given account? Are you dealing with a network of individuals and, if so, what impact does that have on the validity of their information?

Making direct contact will let you build trust and transparency with a social media source. Be open when communicating with sources and give information about your organization and intention.

The nature of the language while reaching out could vary from country to country, due to the sensitivity of the political conflict in the region.   

For example, when working on content from inside Syria, Storyful’s journalists were aware that the conflict was highly complex and sensitive and that Syrian society was polarized. Before reaching out to any party, we asked these questions:

1. Who is the uploader?
2. Who do they support?
3. Are they a media activist group or an individual?
4. Where are they based in Syria?
5. Are they operating in government areas or opposition areas?
6. What is their religious and cultural background?
7. What is their political stance? Are they pro-opposition, pro-government or neutral?

Content from militias, military forces and government sources can be considered public – provided for dissemination. For all other sources, seek permission before using their content.

Use neutral language in all communications with sources on the ground. For example, avoid using the word “militia” when describing rebels or opposition groups when speaking to uploaders who take a pro-opposition stance. The same should be applied while communicating with uploaders who take a pro-government stance. For example, avoid using the word “regime” when describing a government.  

Once a set of reliable accounts has been established, mine them for daily leads and content.

Make use of existing, publicly available sources wherever possible, and work to independently verify information from those sources.

Find relevant lists on Twitter as follows by searching online:

• site:twitter.com/*/lists/keyword
• site:twitter.com/*/lists/*keyword

Account and Search Monitoring

There are multiple methods for list building and alerting across various platforms.

• A list of useful YouTube channels can be created in subscriptions. A subscription list can be exported as an RSS feed.
• Useful Twitter accounts can be added to a list, or a series of lists, and monitored manually or with tools such as Tweetdeck and Hootsuite.

Build systems using free tools to get alerts on content from multiple sources:

• Crowdtangle: Used to build lists of content from sources on Facebook (pages or groups), Instagram, Twitter (accounts or lists) and subreddits
  • Powerful for monitoring all content or for watching for spikes in activity
  • Set up email or Slack alerts from inside the app
• Feedly (or another RSS reader): Used to build lists from YouTube subscriptions, from websites or from audio feeds.
  • Used to monitor news feeds and subscriptions.
• Google Alerts: Set up alerts for search terms (keyword), for instances of keywords on a given site (site:website.com keyword), and for handles of accounts for sources you have already developed.

Centralise alerts from various platforms via email alerts, Slack integrations, etc, based on the needs of the team or project.

Manual searches

Using your relevant source lists, scan for content, stories and topics. From these, develop relevant keywords: locations, terms, topics, and colloquialisms. You should build a profile of what you’re looking for – the who, what, where, when. This data set will differ for theme and topic, but once established give you a starting point for searches.

Use these terms to search for video on the known, most-used platforms, for content and for relevant reporting.

This process should be systematised — conduct a sweep of sources each day or multiple times each day to ensure all relevant leads and content are captured for analysis or archiving.

It is key to understand what people are saying and how they are saying it. Be specific with keywords:

• Don’t use words or phrases that are too common, this will create noise
• Whenever you can, use the street names, building names, and other terms used locally
• Are religious phrases used?
• Are derogatory terms used for either victims or perpetrators?
• Are common misspellings or slang terms used?
• Understanding what hashtags are being used is key to tracking stories on certain platforms, but drop the # symbol when searching manually across platforms — this increases the range of results.
• Platforms have different idiosyncrasies. Instagram concatenates keywords into one search term, for example, so understanding platform limitations is key to tailoring searches. On Instagram, a keyword search for the phrase “London Eye”, for example, will return results for the concatenated term “LondonEye”, rather than an AND search for the two keywords separately. In this case, then, it would be better to search for content geotagged to the London Eye’s location, or to consider alternative keywords.
• Use Boolean operators or advanced search tools on each platform, where available, to get the best results.

Create your own index of useful terms with associated translations. In order to share terms and replicate searches systematically across a team or over time, log terms, translations, slang terms, common misspellings, etc, in a spreadsheet, or build logs of saved searches on platforms such as Twitter and Google for easy replication.

Daesh or Daish (داعش), for example, the Arabic equivalent to the acronym ISIS, is a term that is widely used in the MENA region as a descriptor for Islamic State. The term has negative connotations in Arabic. Therefore, IS media arms and supporters will never use this term.

Different platforms are preferred in different regions, and serve as a good starting point for manual searches, though, of course, there are exceptions. The following guide may prove useful:

Region	Primary Social Media Platforms
Syria, Lebanon, Jordan, Palestine, Iraq, Israel, Yemen	Facebook YouTube Twitter Instagram
Saudi Arabia, United Arab Emirates, Qatar, Kuwait, Bahrain, Oman	Twitter Instagram YouTube Facebook
Egypt, Sudan, Libya, Algeria, Morocco, Tunisia, Mauritania, Western Sahara, Somalia, Djibouti, Comoros	Facebook YouTube Twitter Instagram

This methodology will return leads and content worthy of further investigation.

Finding and Recording Metadata

Applying the leads and content worthy of further examination.

Your results will constantly improve through a circular process that involves: starting with reliable sources for leads; methodical searches of key platforms based on an understanding of the preferred means of communication; expanding, curating and perfecting those searches based on new information gained throughout the process.

From there, seek original content and corroborating information to get as close to the sources and the story as possible.

Systematically logging metadata associated with a piece of content should be your goal. This metadata will vary based on how close an investigation can get to the original of any piece of content.

An original photo or video sourced directly from the owner, or via an archiving or FTP website, may contain original technical metadata about the camera, phone, time and location. Though this should not be trusted implicitly, it may provide some useful starting point for verification.

There are websites and tools that can provide the metadata for some videos and photos. Source the original file to get accurate metadata.

1. Amnesty International DataViewer
2. InVID: YouTube
3. GooFile: Video
4. FotoForensics  
5. Jeffrey’s Image Metadata Viewer
6. Get Metadata
7. Media Info

Perpetrators often disseminate content on social media and, as such, the content will be stripped of original technical information, but will gain metadata relating to upload time, date and location as well as information about the source that can be useful in the verification process.

You can develop a third layer of metadata during the content-verification process, establishing relevant information about the source, date and location of a given piece of content.

A schema for the archiving of content and associated metadata is set out below.

The Verification Process

Storyful’s verification process seeks to establish the original source, date and location for each piece of content using all available open source tools and techniques. Storyful categorises content in terms of three levels of verification: Checking, Corroborated and Confirmed.

The verification process can be more clearly seen in the case studies below.

Date

For any piece of content, sort search results chronologically and replicate the keyword searches across platforms – Do you have the earliest instance of the video?

Use reverse image search on thumbnails, key frames, on-screen logos, etc. Look for earlier versions of the video, or old reports about a similar incident from which the video may be taken.

Do reverse image search or analyse image metadata with:

• Google Image Search – on the web or use Chrome extension
• RevEye – Chrome extension allowing cross-platform image searching
• TinEye
• Jeffrey’s Image Metadata Viewer

You can review metadata on a social video with:

• Amnesty International DataViewer
• InVid

Consider also the number of views, the upload time, the quality of the footage, and the uploader’s username and profile picture. Are the contents, the uploader, the quality consistent with what you’d expect?

If there are no earlier versions, look for hints about when the video was taken: do the weather, time of day, comments made by people in the video, etc, tally with the stated date of the clip? Does the content of the media match reports from independent, trusted sources?

Source

If the answers to the above questions are yes, this may be the earliest source. Are there other ways to find out more about the uploader and people seen in the video?

Review the uploader’s social profile and check for a presence on other channels. Is their publicly available information consistent with the video? Where does the person live? Are they travelling? Is it plausible that they would be at the scene of this incident? Can they be contacted easily? Can they provide an original, raw version of the video?

Use open source tools to monitor their presence online:

• Foller.me – Identify a Twitter user’s friends/associates.
• Graph.tips – Search an individual’s Facebook interactions based on their username.
• Inteltechniques.com – Also for Facebook. See the events a profile has RSVP’d to, places they recommend, places checked into, etc.

In the case of perpetrator video in particular, where an uploader may not want to be identified, what can you learn from the video itself about the uploader or other people seen in the video?

Search for more content that shows the same incident, as it will give a wider perspective. This is central to verifying the source, date and location of the video.

For example, while investigating the suspected chemical attack on the Syrian city of Douma in April 2018, Storyful used multiple videos and photos from the incident to compare and crosscheck the scenes and the information provided to get clearer perspective on the attack, the victims and the perpetrators.

The following independent sources shared videos from the site:

1. The White Helmets
2. Asaad Hanna
3. Yaser Fawal
4. Bilal Abu Salah

Look at military uniforms, weapons, vehicles or munitions seen in the footage to assess the people seen in the video. Use open source information, such as this site on weapon types used in the Syrian conflict. However, information on social media, posted by locals or members of the military, should also be sourced to get up-to-date information from the locality.

If you trace the earliest version of a video back to an archiving site or a closed platform, what can this tell you? Can you find the original post, can you identify the account that uploaded it? Is there an observable spike in conversation, or was the video apparently shared spontaneously by a number of people at around the same time? This may indicate a coordinated campaign.

Location

Verifying the location of a video depends on a careful assessment of unique identifiers: geographic, cultural, clothing, language used, and so on.

Each of these can provide contextual clues to the location or corroborating information.

Search for useful information (street signs, car license plates, buildings, street furniture, etc) that can help you pinpoint an exact location, through search or mapping.

Make use of freely available mapping tools, Wikimapia, Yandex, Baidu, Google Maps, and Bing Maps.

If there are landmarks (such as a mosque, a water tower), filter by building type on Wikimapia:

This will allow you to narrow your search on mapping services that provide satellite imagery or street-level images.

Use advanced techniques to corroborate information in the video – what can you tell from accents, from military uniforms, from weapons used, from vehicles or munitions?

The two case studies appended to this report give numerous examples of geographic verification.

Audience Engagement

The reach of a given video cannot be exhaustively tracked. Views, interactions, etc, will rarely give an accurate picture in the case of perpetrator video, which is often copied, shared and reshared multiple times.

Due to the nature of perpetrator content, platforms often move quickly to remove videos and close accounts. It is, therefore, difficult to track the reach of originals.

Social data can be used, however, to gain an understanding of the target audience of a piece of content and whether that audience is being reached. It may also be possible to identify key amplifiers of a given video.

Consider the Crowdtangle Chrome extension, which tracks shares of a given link across Facebook, Twitter, Instagram and Reddit.

Other services, such as Buzzsumo’s paid social monitoring service, can be used to do more involved metrics analysis.

This sort of analysis is primarily useful in feeding back into building lists of reliable sources of information.

Aggregation of Sensitive Data & Ethical Use

Verified content should be archived systematically.

Download everything. Due to the problematic nature of perpetrator content, videos should be downloaded  immediately, since they may be removed by platforms, or an uploader’s account may be suspended for breach of terms of service – even if the content is uploaded for journalistic or human rights purposes.

Archive everything. The original post, if the content originated on social media, should be archived using a publicly available archiving tool, such as:

• Archive.is
• Archive.org
• Wayback Machine extension/add-on for Archive.org (Chrome, Firefox)

This will maintain metadata of the post (upload date, etc).

However, while it is recommended that you manually back up perpetrator videos or photos due to the likelihood that they will be removed, Storyful does not advise that this content is stored on publicly accessible archiving sites. This may inadvertently allow for the further dissemination of the content.

Screenshot everything. Screenshot posts, comments, and profiles and log the time that screenshots were taken.

Archiving content and associated metadata

Consideration should be given to the nature of the content. It may need to be hosted offline, ideally backed up to multiple drives, or on private servers, as hosting companies may take issue with such content. However, encrypted online storage options are available and many services provide anonymity and privacy. Consider which service is best for your work.

Such content should be archived in a manner that allows it to be retrieved and linked to other content from the same incident as well as with any associated metadata.

Consider the following schema:

• An ID for each incident (can be shared by multiple pieces of content)
• A unique ID for each piece of content
• A unique url for the archived version of the video
• Time/Date data:
  • Time/Date acquired
  • Time/Date uploaded
  • Time/Date of incident – from verification process
• Location data:
  • Location from metadata
  • Location from description
  • Location – from verification process
  • Geo-coordinates (where location has been confirmed through verification process)
• Source data:
  • Uploader name
  • Uploader account url
  • Original source name – from verification process
• A link to archived information (online archive files, screenshots,etc.)
• Discovery data
  • Keywords, hashtags, caption, description
• Ancillary data
  • Platform, video quality, shares, other versions, audience engagement data (where available).

Ethical use of perpetrator content

Collecting and analysing perpetrator content raises several ethical concerns, both in terms of the potential impact on journalists or investigators, and because of concerns relating to the use and dissemination of such material and the risks to the safety and dignity of those involved. This is particularly important for journalists or archivists intending to report on or catalogue human rights abuses.

Consideration to the journalist, analysts or investigators involved in a project:

• Are individuals prepared for the kinds of content they’ll be exposed to?
• Are individuals given adequate support – training, counselling, etc?
• Is the workflow such that an individual can step back from the work without impacting the project?
• Is the team staffed up to allow members alternate between tasks, or to ensure no one person is overexposed to difficult content.  

Consideration should be given to how to store and archive the content:

• As discussed above, sensitive content should not be archived on public platforms or shared for any purpose not central to the project or investigation.
• Storing sensitive content on video platforms or commercial servers risks the content being shared and may raise problems from service providers.

Consideration should be given to the victims and individuals involved:

• In the case of perpetrator video, the content will have been published intentionally for public dissemination. As such, the original source’s consent is not required for publication; however, if victims can be identified, it should be considered whether they should be contacted before publication.
• Will publishing or making available the content endanger anyone or injure the dignity of those involved?
• Will editing the content to protect victims, sources or individuals impact the validity or verifiability of the content?
• Can the content be archived in such a way that it is searchable and available as necessary, without publication.

Consideration should also be given to the target audience:

• The target audience could be members of the public, journalists, or legal professionals.
• Graphic content should be identifiable and end users should have control over what they’re exposed to when receiving this sort of information.

Before publishing or disseminating the content, it must be verified and archived in a manner that will give the end user all the information they need to take action on the content in an informed manner. This should include information that will ensure the target audience avoids creating problems for the victims, their families, or their wider community through publication or dissemination.

Consideration must be given to whether publishing or using the content will inadvertently fuel tension or possibly lead to an escalation of violence, or risk to the victims.

Storyful’s journalists approach these stories cautiously, providing partner newsrooms with the information they need to tell the stories accurately and confidently in the style and format that meets their own editorial and ethical guidelines.

The verification processes set out in this report is applicable to newsrooms and investigators and is designed to prevent misreporting or wrongful use of content of this type.

Storyful’s policy is never to publish perpetrator content publicly (though we archive and provide such content to partners).

In order to protect the integrity of our work, Storyful considers offering anonymity to sources only where there is a compelling case to do so, such a potential threat to life, liberty or security.

Storyful has specific policies around the publication of content that shows minors (parental releases may be required), or where the blurring of faces, muting audio to protect sensitive information, etc, may be necessary. In our role as a newsroom-to-newsroom service provider, our partners are the ultimate arbiters of how this content will be used.

Storyful cannot provide legal advice on these matters to newsrooms or investigators. Individual newsrooms will be bound by local legislation with regard to the right to a free trial, privacy and other concerns. Storyful’s role is to provide detailed information to our partners to allow them to make informed decisions about usage.

Risk Assessment

The goal of publication of content by journalists or human rights organizations is to draw attention to an issue, or to prompt action by authorities to deal with the situation or bring perpetrators to justice. Consider the following:

Public Interest vs Individual Risk

• Will the publication serve the public interest, but put an individual at risk? Storyful would not advise that content be published in such cases.
• Is it possible to tell the story in a different format, or, for human rights investigators, can the case be handed over to authorities for investigation without publication?
• If there is a strong public interest case for publication, can the uploader or the victim be identified? If so, can their identity be obscured while still allowing their story to be told?

Public Interest vs Rights of the Accused

• Will the publication serve to draw attention to an issue, but put prosecution at risk where the perpetrator is identified?
• In the case of a high-profile perpetrator, there is a case to be made for the publication to force the authorities to take action. The case of Mahmoud al-Werfalli, a Libyan special forces commander wanted by the International Criminal Court, is an example where social media content, in which al-Werfalli was clearly identifiable, was central to the prosecution.
• There is a strong case for publication in the public interest where the content is verified and the source is a public figure, if the goal is to draw attention to the issue or the crime.
• However, the rights of the accused also need to be protected. Depending on the case and the circumstances, blur faces or take other steps to insure the story can be told without prejudicing prosecution.

Public Interest vs Individual Dignity

• Will publishing a piece of content result in greater advocacy or better long-term results for a victim than would otherwise be the case, even where publication would impact the person’s privacy or dignity?
• Special consideration is given to the following cases:

Children: Children are highly vulnerable, so respecting their dignity and privacy is vital. It is important to obtain permission and consent from parents for all photos and videos used for publication.

UNICEF’s guidelines for journalists reporting on children recommend to “always change the name and obscure the visual identity of any child” that meets the following criteria:

• Is a victim of sexual abuse or exploitation
• Is a perpetrator of physical or sexual abuse
• Is HIV positive, or living with AIDS, unless the child, a parent or a guardian gives fully informed consent
• Is charged or convicted of a crime

Women: Considerations around women and their safety are particularly relevant in certain regions.

As the status of women varies widely from city to city and country to country, special considerations should be taken when reporting on such cases of abuse.

In some places, a woman’s “honor” can be important to her family or society, increasing the risk that identifying a female source or victim could lead to reprisals or put such people at risk of death. Women can become “double victims”: they may be victims of the perpetrator, as well as victims of the tradition and rules of their society, or may become victimised as a result of being the subject of reporting or an investigation.

The fear of being subjected to blame and public criticism if they are named in the media can lead people to decide against speaking up about their experiences. It is important that the use or publication of perpetrator content does not add to the factors that constrain a victim’s ability to speak out.   

Prisoners: Videos of prisoners often show them in vulnerable situations. Perpetrator-created videos sometimes show prisoners being humiliated, degraded or even tortured.

Videos of this kind are often shared with the intent to warn a community of the consequences of disobeying orders, or as acts of dominance over a defeated enemy.

Perpetrators may want to start negotiations with the families of the prisoners, as set out in the first case study below.

Specific cases, such as that of the IS-captured British journalist John Cantlie, kidnapped in Syria in November 2012, raise specific questions. Cantlie appeared in a number of propaganda videos, one of which is discussed in the second case study appended to this report.

Storyful’s policy is never to publish directly content relating to Cantlie; however, we have archived, verified, and distributed his work to our partners. Newsrooms reporting on these video releases often use screenshots or even video clips, in order to illustrate Cantlie’s statements or to report on his location.

Considerations for Survivors of Sexual Abuse

Victims identified in perpetrator content may have been subjected to sexual abuse or exploitation.

Consider the following points in dealing with content in these cases:  

1. Anonymity (unless the person chooses otherwise after getting advice from a legal advisor).
2. Only adults can make an informed decision about how they want their identity to be reported.
3. Children’s identity should not be reported, even if they choose not to remain anonymous.
4. Don’t share any information that, taken as a whole, could lead others to identify the survivors, such as names, personal details, nationality, hometown, or age.
5. Use non-judgemental language in reporting and while conducting interviews or outreach.
6. Maintain dignity, confidentiality and respect.

Where appropriate, in each of the cases outlined above, consider not providing the full name of the victims, blurring content for public distribution, etc, while carefully documenting the facts for archival purposes.  

The Dart Center for Journalism and Trauma spoke to journalists and trauma experts in 2016, who provided useful information for journalists, on both self care and on working with and speaking to the victims of child sexual abuse.