2017 was undoubtedly a most interesting year for politics and media. POTUS Donald J. Trump continued to make all kinds (often the wrong kind) of headlines for his tirades against inaccurate journalism and misinformation. Across the world in India too, misinformation and inaccurate (and quite often even maliciously untruthful) journalism started to impact elections, popular morale and the effectiveness of governmental initiatives. "Fake News" hence became such a household term around the world that U.K. based Collins Dictionary chose "Fake News" as its 2017 Word of the Year.
“Fake news, either as a statement of fact or as an accusation, has been inescapable this year, contributing to the undermining of society’s trust in news reporting,” said Helen Newstead, Collins’ head of language content.
While the term has indeed been popularized by politicians and statesman over the last 2 years, fake news has of course been around (and been a problem) for as long as we have had politics, media, warfare or civilization itself.
The great Indian epic Mahabharata tells a war story where the Pandavas willfully propagated a fake news of the warrior Ashwathama's death (when it was in fact an elephant also named Ashwathama who had been killed) in order to rattle and ultimately defeat the opposing general Dronacharya of the Kauravas.
In ancient Egypt, though his architectural achievements were remarkable on their own, Pharaoh Ramses II and his supporters greatly exaggerated his military achievements when documenting his reign (1279-1213 BCE) on inscriptions – leading later historians to misunderstand and “over-assume” the course of Egyptian westward expansion into Libya and surrounding areas.
Roman emperor Augustus spread fake news to malign Mark Antony (alleging Antony had denounced his Roman religion, ethos and heritage in favor of an Egyptian one and even intended to be Ptolemized as a Pharaoh) in the eyes of Roman society - and popularize the assault against Cleopatra.
Early Christian history too is abundant in stories of the Catholic Church propagating fake news about practices of pagans, Jews and even other opposing schools of Christian thought. The early Church in fact fabricated a document now known as the “Donation of Constantine” alleging that Emperor Constantine had transferred land and political control to the Pope and used the document to justify the Church’s political, administrative, fiscal and moral control over large tracts of land. It was similarly not uncommon in non-Christian societies to hear of barbaric Christian rituals, including priestly cannibalism, necrophilia and sodomy.
Rulers around the world from antiquity to the modern era, have relied on fake news and propaganda as an important tool for garnering popular (and financial) support for their military and political campaigns. Benjamin Franklin was guilty of fabricating a whole newspaper worth of propaganda alleging, with gruesome detail, Native American support for the English crown.
In World War 1, the Allied Powers popularized the notion of a German Corpse Factory (Kadaververwertungsanstalt) that Germans supposedly used to turn dead bodies from warzones to nitroglycerine, candles, lubricants, and boot dubbin.
In World War 2, Great Britain instituted the Political Warfare Executive (PWE) for creating and spreading false anti-German propaganda to further mobilize their citizens against the Axis powers. In much the same way fake news spreads on the Internet today, the PWE delivered fake propaganda (and even real-time reporting of non-existent bombing raids) embedded within reliable news and information on events in Germany and other German-occupied areas, making it impossible to the public (and in some cases, even historians today) to determine what was real and what was fabricated. The media in Axis-held areas similarly also disseminated fake news about Jews in particular, societies and politics of the Allied powers in general, as well as military campaigns throughout the war in much the same way.
Fake news has no less impact in our political and social life today. With the advent and proliferation of new media (read the Internet), it has become increasingly easy for agenda-driven individuals and institutions to propagate fake news. All it takes is one piece of false content to be placed in the right forum and the human need to disperse and share sensational information can make it the accepted truth overnight. The key inflexion point in the process is the choosing or being the "right forum" - which is where even institutions become willing or unwilling perpetrators. With online advertising, it has also become increasingly profitable for agenda-driven institutions of our media to become the "right forum" - a media institution for example is inherently incentivized to publish fake content since it will increase inbound traffic (and thus advertising revenues). In a 24 hour news environment, the audience tends to have a short enough memory span to forget (and forgive) journalistic lapses even if the fake content is later proven to be fake and/or is simply redacted.
Yet, whether or not you agree with politicians and public discourse on their characterization on what is fake news, fake news and content does irreparable harm to public policy, public safety and general discourse on important issues of our day. It widens disagreements, causes public policy decisions to be impulsive, and some might argue, cause general discontent and stress in the society. At the very least, it distracts the conversation away from issues that really matter - poverty, hunger, inequality and irreparable harm to the environment.
Necessity is of course the mother of invention. Several people have recognized that in the new media at least, while it is algorithmically easier to propagate fake news, it is algorithmically easier to detect and counter fake news as well. See below for just a few examples:
US20170177717A1, “Rating a level of journalistic distortion in news and media content”
Journalistic distortion is the inclusion of distortion in news media content in the form of incorrect facts, support for an agenda or bias,
political or other external influence etc. The patent discloses a method to provide a rating to news articles that indicated the level of distortion in it. It works by selecting the first news article from a plurality of news sources, analyses it via a stored algorithm and calculates a rating according to predefined categories. These categories include incorrect facts, bias, spin, slant, influence etc. The rating is then accumulated for all categories and the final rating is delivered the user’s device who wishes to access the news content. The algorithm uses multiple dictionaries which contain keywords that indicate distortion of some kind.
The analysis is performed in four tiers: individual words, phrases and sentences, paragraphs and the article as a whole. The analysis keeps proceeding onto the next tier in order find the accurate rating. For example, concepts which involve sarcasm or irony may not be detected at an initial tier and may need to be analyzed on a higher tier.
EP2937824A9, “System and method for evaluating the credibility of news emerging in social networks for information and news reporting purposes”
The invention evaluates the quality of news emerging from social networks. Presently, such analysis made on users and the content posted by users over a narrow domain of metrics, which can quantify only a particular category at a time. The system improves upon it and enables analysis over a broad domain of metrics. The metrics apply to both the users of social media services as well as the content written/shared by them. The system is named “AlethioMeter” and includes an NLP engine used to extract sentiment from content available online and a Network Analytics Engine which performs calculations to assign a score to the social media data. In fact, a total of 35 metrics are discussed which fall in the following three pillars.
Contributor: Reputation, History, Popularity, Influence, Presence of source of the post.
Content: Reputation (of links included), History (of links included), Originality (of multimedia included), Authenticity (of multimedia included), Proximity (of multimedia included) in the content of the post.
Context: Co-occurrences about the same thing on other sources, internal and external coherence with tags, attached links and multimedia, location from where the post was written compared to the location mentioned on the post itself.
Finally, an index value is computed from all the metrics that quantifies the quality of the content being scrutinized.
US9186514, “Optimized Fact Checking Method and System”
The patent teaches a fact checking system that determines the factual validity, accuracy and quality of an article through context comparison, pattern matching and natural language comparison. It uses lexical chaining to summarize the article and utilizes social network information of the user to determine the focus of the article and predefined templates in order to summarize the article and fact check the article, the summary or both.
CA2984904A1, “Social Media Events Detection and Verification”
The patent describes a method to detect and verify social media events, wherein useful information is received from social media data via an Event Detecting Server (EDS). The key components of the EDS include modules for Ingestion, Filtering, Organization, Clustering, Verification, Categorization, Summarization, News-Worthiness, Opinion and Credibility.
The Ingestion module retrieves the information from a social media platform and stores it with associated metadata into an ingested data store of EPS.
The Filtering module based on language and profanity removes the inappropriate data. It may also use a classification algorithm to remove data classified as spam, chat or advertisements.
The Organization module then fetches the filtered data from filtered data store to determine key concepts which are then organised into a database by the Clustering module.
The Verification module then determines the level of accuracy of the event detected cluster by generating a veracity calculation based on: user, tweet level or social media data level.
The Categorization module categorizes the data collected by event cluster module in the topics such as sports, entertainment, business, finance etc.
The Summarization module selects a unit of data, based on metrics such as popular unit data, to be added as a summary to the metadata for particular event detected cluster.
The News-Worthiness module uses Machine Learning algorithm to generate newsworthiness score.
The Opinion module detects if the each unit data contains opinion of a person or assertion of fact.
The Credibility module generates confidence score based on three components: source credibility, cluster credibility and tweet credibility. Source Credibility determine the authentication of source, Cluster Credibility determines whether information is genuine or fake based on historic data and Tweet Credibility relates to the contents of individual tweets.
US 2017/0195125, “Promoting Learned Discourse in Online Media with Consideration of Sources and Provenance”
The invention evaluates the authenticity of news or comments posted on a forum server and awarding a digital portable certificate for users that are committed and have been well established in writing or posting news on particular forum. A method for hashing, encrypting and digitally sign the comments or news from a well-established user is used to protect the user’s reputation as well as to ease the process of citation of their content by others.
The host site uses the automated content analyst which relies on relevance, uniqueness, and sentiment/ keyword extraction. The host site uses the bibliographic date to retrieve the public key of the indicated source, and uses that public key to verify that the cited material is authentic. Whenever the host site receives a comment or news it is submitted to automated content analysis filter.
To further evaluate the authenticity of news, a veracity content analysis filter checks for the presence of hashed material that can be verified by processing with a public key.
The host site can maintain (or periodically acquire) its own white lists and black lists, and provide warning services to its users (readers and viewers) in addition to inputs for its overall scoring function. Further, various methods may be used to track user performance and reward users that generate high quality comments with enhanced status. Comment quality can be judged according to a priori score, a posteriori score, or the post moderation score.
While any of the above (or even all of the above combined) may not fully be able to rid the public discourse of fake news, these are good beginnings for restoring public confidence in our news and content. Gartner projects that "[by] 2022, the majority of individuals in mature economies will consume more false information than true information". Hopefully for the sake of all of us, Gartner will be proven wrong just this time.