Tag Archives: Credibility

Got TweetCred? Use it To Automatically Identify Credible Tweets (Updated)

Update: Users have created an astounding one million+ tags over the past few weeks, which will help increase the accuracy of TweetCred in coming months as we use these tags to further train our machine learning classifiers. We will be releasing our Firefox plugin in the next few days. In the meantime, we have just released our paper on TweetCred which describes our methodology & classifiers in more detail.

What if there were a way to automatically identify credible tweets during major events like disasters? Sounds rather far-fetched, right? Think again.

The new field of Digital Information Forensics is increasingly making use of Big Data analytics and techniques from artificial intelligence like machine learning to automatically verify social media. This is how my QCRI colleague ChaTo et al. already predicted both credible and non-credible tweets generated after the Chile Earthquake (with an accuracy of 86%). Meanwhile, my colleagues Aditi, et al. from IIIT Delhi also used machine learning to automatically rank the credibility of some 35 million tweets generated during a dozen major international events such as the UK Riots and the Libya Crisis. So we teamed up with Aditi et al. to turn those academic findings into TweetCred, a free app that identifies credible tweets automatically.

CNN TweetCred

We’ve just launched the very first version of TweetCred—key word being first. This means that our new app is still experimental. On the plus side, since TweetCred is powered by machine learning, it will become increasingly accurate over time as more users make use of the app and “teach” it the difference between credible and non-credible tweets. Teaching TweetCred is as simple as a click of the mouse. Take the tweet below, for example.

ARC TweetCred Teach

TweetCred scores each tweet based based on a 7-point system, the higher the number of blue dots, the more credible the content of the tweet is likely to be. Note that a TweetCred score also takes into account any pictures or videos included in a tweet along with the reputation and popularity of the Twitter user. Naturally, TweetCred won’t always get it right, which is where the teaching and machine learning come in. The above tweet from the American Red Cross is more credible than three dots would suggest. So you simply hover your mouse over the blue dots and click on the “thumbs down” icon to tell TweetCred it got that tweet wrong. The app will then ask you to tag the correct level of credibility for that tweet is.

ARC TweetCred Teach 3

That’s all there is to it. As noted above, this is just the first version of TweetCred. The more all of us use (and teach) the app, the more accurate it will be. So please try it out and spread the word. You can download the Chrome Extension for TweetCred here. If you don’t use Chrome, you can still use the browser version here although the latter has less functionality. We very much welcome any feedback you may have, so simply post feedback in the comments section below. Keep in mind that TweetCred is specifically designed to rate the credibility of disaster/crisis related tweets rather than any random topic on Twitter.

As I note in my book Digital Humanitarians (forthcoming), empirical studies have shown that we’re less likely to spread rumors on Twitter if false tweets are publicly identified by Twitter users as being non-credible. In fact, these studies show that such public exposure increases the number of Twitter users who then seek to stop the spread of said of rumor-related tweets by 150%. But, it makes a big difference whether one sees the rumors first or the tweets dismissing said rumors first. So my hope is that TweetCred will help accelerate Twitter’s self-correcting behavior by automatically identifying credible tweets while countering rumor-related tweets in real-time.

This project is a joint collaboration between IIIT and QCRI. Big thanks to Aditi and team for their heavy lifting on the coding of TweetCred. If the experiments go well, my QCRI colleagues and I may integrate TweetCred within our AIDR (Artificial Intelligence for Disaster Response) and Verily platforms.

Bio

See also:

  • New Insights on How to Verify Social Media [link]
  • Predicting the Credibility of Disaster Tweets Automatically [link]
  • Auto-Ranking Credibility of Tweets During Major Events [link]
  • Auto-Identifying Fake Images on Twitter During Disasters [link]
  • Truth in the Age of Social Media: A Big Data Challenge [link]
  • Analyzing Fake Content on Twitter During Boston Bombings [link]
  • How to Verify Crowdsourced Information from Social Media [link]
  • Crowdsourcing Critical Thinking to Verify Social Media [link]
  • Tweets, Crises and Behavioral Psychology: On Credibility and Information Sharing [link]

New Insights on How To Verify Social Media

The “field” of information forensics has seen some interesting developments in recent weeks. Take the Verification Handbook or Twitter Lie-Detector project, for example. The Social Sensor project is yet another new initiative. In this blog post, I seek to make sense of these new developments and to identify where this new field may be going. In so doing, I highlight key insights from each initiative. 

VHandbook1

The co-editors of the Verification Handbook remind us that misinformation and rumors are hardly new during disasters. Chapter 1 opens with the following account from 1934:

“After an 8.1 magnitude earthquake struck northern India, it wasn’t long before word circulated that 4,000 buildings had collapsed in one city, causing ‘innumerable deaths.’ Other reports said a college’s main building, and that of the region’s High Court, had also collapsed.”

These turned out to be false rumors. The BBC’s User Generated Content (UGC) Hub would have been able to debunk these rumors. In their opinion, “The business of verifying and debunking content from the public relies far more on journalistic hunches than snazzy technology.” So they would have been right at home in the technology landscape of 1934. To be sure, they contend that “one does not need to be an IT expert or have special equipment to ask and answer the fundamental questions used to judge whether a scene is staged or not.” In any event, the BBC does not “verify something unless [they] speak to the person that created it, in most cases.” What about the other cases? How many of those cases are there? And how did they ultimately decide on whether the information was true or false even though they did not  speak to the person that created it?  

As this new study argues, big news organizations like the BBC aim to contact the original authors of user generated content (UGC) not only to try and “protect their editorial integrity but also because rights and payments for newsworthy footage are increasingly factors. By 2013, the volume of material and speed with which they were able to verify it [UGC] were becoming significant frustrations and, in most cases, smaller news organizations simply don’t have the manpower to carry out these checks” (Schifferes et al., 2014).

Credit: ZDnet

Chapter 3 of the Handbook notes that the BBC’s UGC Hub began operations in early 2005. At the time, “they were reliant on people sending content to one central email address. At that point, Facebook had just over 5 million users, rather than the more than one billion today. YouTube and Twitter hadn’t launched.” Today, more than 100 hours of content is uploaded to YouTube every minute; over 400 million tweets are sent each day and over 1 million pieces of content are posted to Facebook every 30 seconds. Now, as this third chapter rightly notes, “No technology can automatically verify a piece of UGC with 100 percent certainty. However, the human eye or traditional investigations aren’t enough either. It’s the combination of the two.” New York Times journalists concur: “There is a problem with scale… We need algorithms to take more onus off human beings, to pick and understand the best elements” (cited in Schifferes et al., 2014).

People often (mistakenly) see “verification as a simple yes/no action: Something has been verified or not. In practice, […] verification is a process” (Chapter 3). More specifically, this process is one of satisficing. As colleagues Leysia Palen et al.  note in this study, “Information processing during mass emergency can only satisfice because […] the ‘complexity of the environment is immensely greater than the computational powers of the adaptive system.’” To this end, “It is an illusion to believe that anyone has perfectly accurate information in mass emergency and disaster situations to account for the whole event. If someone did, then the situation would not be a disaster or crisis.” This explains why Leysia et al seek to shift the debate to one focused on the helpfulness of information rather the problematic true/false dichotomy.

Credit: Ann Wuyts

“In highly contextualized situations where time is of the essence, people need support to consider the content across multiple sources of information. In the online arena, this means assessing the credibility and content of information distributed across [the web]” (Leysia et al., 2011). This means that, “Technical support can go a long way to help collate and inject metadata that make explicit many of the inferences that the every day analyst must make to assess credibility and therefore helpfulness” (Leysia et al., 2011). In sum, the human versus computer debate vis-a-vis the verification of social media is somewhat pointless. The challenge moving forward resides in identifying the best ways to combine human cognition with machine computing. As Leysia et al. rightly note, “It is not the job of the […] tools to make decisions but rather to allow their users to reach a decision as quickly and confidently as possible.”

This may explain why Chapter 7 (which I authored) applies both human and advanced computing techniques to the verification challenge. Indeed, I explicitly advocate for a hybrid approach. In contrast, the Twitter Lie-Detector project known as Pheme apparently seeks to use machine learning alone to automatically verify online rumors as they spread on social networks. Overall, this is great news—the more groups that focus on this verification challenge, the better for those us engaged in digital humanitarian response. It remains to be seen, however, whether machine learning alone will make Pheme a success.

pheme

In the meantime, the EU’s Social Sensor project is developing new software tools to help journalists assess the reliability of social media content (Schifferes et al., 2014). A preliminary series of interviews revealed that journalists were most interested in Social Sensor software for:

1. Predicting or alerting breaking news

2. Verifying social media content–quickly identifying who has posted a tweet or video and establishing “truth or lie”

So the Social Sensor project is developing an “Alethiometer” (Alethia is Greek for ‘truth’) to “meter the credibility of of information coming from any source by examining the three Cs—Contributors, Content and Context. These seek to measure three key dimensions of credibility: the reliability of contributors, the nature of the content, and the context in which the information is presented. This reflects the range of considerations that working journalists take into account when trying to verify social media content. Each of these will be measured by multiple metrics based on our research into the steps that journalists go through manually. The results of [these] steps can be weighed and combined [metadata] to provide a sense of credibility to guide journalists” (Schifferes et al., 2014).

SocialSensor1

On our end, my colleagues and at QCRI are continuing to collaborate with several partners to experiment with advanced computing methods to address the social media verification challenge. As noted in Chapter 7, Verily, a platform that combines time-critical crowdsourcing and critical thinking, is still in the works. We’re also continuing our collaboration on a Twitter credibility plugin (more in Chapter 7). In addition, we are exploring whether we can microtask the computation of source credibility scores using MicroMappers.

Of course, the above will sound like “snazzy technologies” to seasoned journalists with no background or interest in advanced computing. But this doesn’t seem to stop them from complaining that “Twitter search is very hit and miss;” that what Twitter “produces is not comprehensive and the filters are not comprehensive enough” (BBC social media expert, cited in Schifferes et al., 2014). As one of my PhD dissertation advisors (Clay Shirky) noted a while back already, information overflow (Big Data) is due to “Filter Failure”. This is precisely why my colleagues and I are spending so much of our time developing better filters—filters powered by human and machine computing, such as AIDR. These types of filters can scale. BBC journalists on their own do not, unfortunately. But they can act on hunches and intuition based on years of hands-on professional experience.

The “field” of digital information forensics has come along way since I first wrote about how to verify social media content back in 2011. While I won’t touch on the Handbook’s many other chapters here, the entire report is an absolute must read for anyone interested and/or working in the verification space. At the very least, have a look at Chapter 9, which combines each chapter’s verification strategies in the form of a simple check-list. Also, Chapter 10 includes a list of  tools to aid in the verification process.

In the meantime, I really hope that we end the pointless debate about human versus machine. This is not an either/or issue. As a colleague once noted, what we really need is a way to combine the power of algorithms and the wisdom of the crowd with the instincts of experts.

bio

See also:

  • Predicting the Credibility of Disaster Tweets Automatically [link]
  • Auto-Ranking Credibility of Tweets During Major Events [link]
  • Auto-Identifying Fake Images on Twitter During Disasters [link]
  • Truth in the Age of Social Media: A Big Data Challenge [link]
  • Analyzing Fake Content on Twitter During Boston Bombings [link]
  • How to Verify Crowdsourced Information from Social Media [link]
  • Crowdsourcing Critical Thinking to Verify Social Media [link]

Using Crowdsourcing to Counter the Spread of False Rumors on Social Media During Crises

My new colleague Professor Yasuaki Sakamoto at the Stevens Institute of Tech-nology (SIT) has been carrying out intriguing research on the spread of rumors via social media, particularly on Twitter and during crises. In his latest research, “Toward a Social-Technological System that Inactivates False Rumors through the Critical Thinking of Crowds,” Yasu uses behavioral psychology to under-stand why exposure to public criticism changes rumor-spreading behavior on Twitter during disasters. This fascinating research builds very nicely on the excellent work carried out by my QCRI colleague ChaTo who used this “criticism dynamic” to show that the credibility of tweets can be predicted (by topic) with-out analyzing their content. Yasu’s study also seeks to find the psychological basis for the Twitter’s self-correcting behavior identified by ChaTo and also John Herman who described Twitter as a  “Truth Machine” during Hurricane Sandy.

criticalthink

Twitter is still a relatively new platform, but the existence and spread of false rumors is certainly not. In fact, a very interesting study dated 1950 found that “in the past 1,000 years the same types of rumors related to earthquakes appear again and again in different locations.” Early academic studies on the spread of rumors revealed that “that psychological factors, such as accuracy, anxiety, and impor-tance of rumors, affect rumor transmission.” One such study proposed that the spread of a rumor “will vary with the importance of the subject to the individuals concerned times the ambiguity of the evidence pertaining to the topic at issue.” Later studies added “anxiety as another key element in rumormongering,” since “the likelihood of sharing a rumor was related to how anxious the rumor made people feel. At the same time, however, the literature also reveals that counter-measures do exist. Critical thinking, for example, decreases the spread of rumors. The literature defines critical thinking as “reasonable reflective thinking focused on deciding what to believe or do.”

“Given the growing use and participatory nature of social media, critical thinking is considered an important element of media literacy that individuals in a society should possess.” Indeed, while social media can “help people make sense of their situation during a disaster, social media can also become a rumor mill and create social problems.” As discussed above, psychological factors can influence rumor spreading, particularly when experiencing stress and mental pressure following a disaster. Recent studies have also corroborated this finding, confirming that “differences in people’s critical thinking ability […] contributed to the rumor behavior.” So Yasu and his team ask the following interesting question: can critical thinking be crowdsourced?

Screen Shot 2013-03-30 at 3.37.40 PM

“Not everyone needs to be a critical thinker all the time,” writes Yasu et al. As long as some individuals are good critical thinkers in a specific domain, their timely criticisms can result in an emergent critical thinking social system that can mitigate the spread of false information. This goes to the heart of the self-correcting behavior often observed on social media and Twitter in particular. Yasu’s insight also provides a basis for a bounded crowdsourcing approach to disaster response. More on this here, here and here.

“Related to critical thinking, a number of studies have paid attention to the role of denial or rebuttal messages in impeding the transmission of rumor.” This is the more “visible” dynamic behind the self-correcting behavior observed on Twitter during disasters. So while some may spread false rumors, others often try to counter this spread by posting tweets criticizing rumor-tweets directly. The following questions thus naturally arise: “Are criticisms on Twitter effective in mitigating the spread of false rumors? Can exposure to criticisms minimize the spread of rumors?”

Yasu and his colleagues set out to test the following hypotheses: Exposure to criticisms reduces people’s intent to spread rumors; which mean that ex-posure to criticisms lowers perceived accuracy, anxiety, and importance of rumors. They tested these hypotheses on 87 Japanese undergraduate and grad-uate students by using 20 rumor-tweets related to the 2011 Japan Earthquake and 10 criticism-tweets that criticized the corresponding rumor-tweets. For example:

Rumor-tweet: “Air drop of supplies is not allowed in Japan! I though it has already been done by the Self- Defense Forces. Without it, the isolated people will die! I’m trembling with anger. Please retweet!”

Criticism-tweet: “Air drop of supplies is not prohibited by the law. Please don’t spread rumor. Please see 4-(1)-4-.”

The researchers found that “exposing people to criticisms can reduce their intent to spread rumors that are associated with the criticisms, providing support for the system.” In fact, “Exposure to criticisms increased the proportion of people who stop the spread of rumor-tweets approximately 1.5 times [150%]. This result indicates that whether a receiver is exposed to rumor or criticism first makes a difference in her decision to spread the rumor. Another interpretation of the result is that, even if a receiver is exposed to a number of criticisms, she will benefit less from this exposure when she sees rumors first than when she sees criticisms before rumors.”

Screen Shot 2013-03-30 at 3.53.02 PM

Findings also revealed three psychological factors that were related to the differences in the spread of rumor-tweets: one’s own perception of the tweet’s accuracy, the anxiety cause by the tweet, and the tweet’s perceived importance. The results also indicate that “exposure to criticisms reduces the perceived accuracy of the succeeding rumor-tweets, paralleling the findings by previous research that refutations or denials decrease the degree of belief in rumor.” In addition, the perceived accuracy of criticism-tweets by those exposed to rumors first was significantly higher than the criticism-first group. The results were similar vis-à-vis anxiety. “Seeing criticisms before rumors reduced anxiety associated with rumor-tweets relative to seeing rumors first. This result is also consistent with previous research findings that denial messages reduce anxiety about rumors. Participants in the criticism-first group also perceived rumor-tweets to be less important than those in the rumor-first group.” The same was true vis-à-vis the perceived importance of a tweet. That said, “When the rumor-tweets are perceived as more accurate, the intent to spread the rumor-tweets are stronger; when rumor-tweets cause more anxiety, the intent to spread the rumor-tweets is stronger; when the rumor-tweets are perceived as more im-portance, the intent to spread the rumor-tweets is also stronger.”

So how do we use these findings to enhance the critical thinking of crowds and design crowdsourced verification platforms such as Verily? Ideally, such a platform would connect rumor tweets with criticism-tweets directly. “By this design, information system itself can enhance the critical thinking of the crowds.” That said, the findings clearly show that sequencing matters—that is, being exposed to rumor tweets first vs criticism tweets first makes a big differ-ence vis-à-vis rumor contagion. The purpose of a platform like Verily is to act as a repo-sitory for crowdsourced criticisms and rebuttals; that is, crowdsourced critical thinking. Thus, the majority of Verily users would first be exposed to questions about rumors, such as: “Has the Vincent Thomas Bridge in Los Angeles been destroyed by the Earthquake?” Users would then be exposed to the crowd-sourced criticisms and rebuttals.

In conclusion, the spread of false rumors during disasters will never go away. “It is human nature to transmit rumors under uncertainty.” But social-technological platforms like Verily can provide a repository of critical thinking and ed-ucate users on critical thinking processes themselves. In this way, we may be able to enhance the critical thinking of crowds.


bio

See also:

  • Wiki on Truthiness resources (Link)
  • How to Verify and Counter Rumors in Social Media (Link)
  • Social Media and Life Cycle of Rumors during Crises (Link)
  • How to Verify Crowdsourced Information from Social Media (Link)
  • Analyzing the Veracity of Tweets During a Crisis (Link)
  • Crowdsourcing for Human Rights: Challenges and Opportunities for Information Collection & Verification (Link)
  • The Crowdsourcing Detective: Crisis, Deception and Intrigue in the Twittersphere (Link)

Tweeting is Believing? Analyzing Perceptions of Credibility on Twitter

What factors influence whether or not a tweet is perceived as credible? According to this recent study, users have “difficulty discerning truthfulness based on con-tent alone, with message topic, user name, and user image all impacting judg-ments of tweets and authors to varying degrees regardless of the actual truth-fulness of the item.”

For example, “Features associated with low credibility perceptions were the use of non-standard grammar and punctuation, not replacing the default account image, or using a cartoon or avatar as an account image. Following a large number of users was also associated with lower author credibility, especially when unbalanced in comparison to follower count [...].” As for features enhan-cing a tweet’s credibility, these included “author influence (as measured by follower, retweet, and  mention counts), topical expertise (as established through a Twitter homepage bio, history of on-topic tweeting, pages outside of Twitter, or having a location relevant to the topic of the tweet), and reputation (whether an author is someone a user follows, has heard of, or who has an official Twitter account verification seal). Content related features viewed as credibility-enhancing were containing a URL leading to a high-quality site, and the existence of other tweets conveying similar information.”

 In general, users’ ability to “judge credibility in practice is largely limited to those features visible at-a-glance in current UIs (user picture, user name, and tweet content). Conversely, features that often are obscured in the user interface, such as the bio of a user, receive little attention despite their ability to impact cred-ibility judgments.” The table below compares a features’s perceived credibility impact with the attention actually allotted to assessing that feature.

“Message topic influenced perceptions of tweet credibility, with science tweets receiving a higher mean tweet credibility rating than those about either politics  or entertainment. Message topic had no statistically significant impact on perceptions of author credibility.” In terms of usernames, “Authors with topical names were considered more credible than those with traditional user names, who were in turn considered more credible than those with internet name styles.” In a follow up experiment, the study analyzed perceptions of credibility vis-a-vis a user’s image, i.e., the profile picture associated with a given Twitter account. “Use of the default Twitter icon significantly lowers ratings of content and marginally lowers ratings of authors [...]” in comparison to generic, topical, female and male images.

Obviously, “many of these metrics can be faked to varying extents. Selecting a topical username is trivial for a spam account. Manufacturing a high follower to following ratio or a high number of retweets is more difficult but not impossible. User interface changes that highlight harder to fake factors, such as showing any available relationship between a user’s network and the content in question, should help.” Overall, these results “indicate a discrepancy between features people rate as relevant to determining credibility and those that mainstream social search engines make available.” The authors of the study conclude by suggesting changes in interface design that will enhance a user’s ability to make credibility judgements.

“Firstly, author credentials should be accessible at a glance, since these add value and users rarely take the time to click through to them. Ideally this will include metrics that convey consistency (number of tweets on topic) and legitimization by other users (number of mentions or retweets), as well as details from the author’s Twitter page (bio, location, follower/following counts). Second, for con-tent assessment, metrics on number of retweets or number of times a link has been shared, along with who is retweeting and sharing, will provide consumers with context for assessing credibility. [...] seeing clusters of tweets that conveyed similar messages was reassuring to users; displaying such similar clusters runs counter to the current tendency for search engines to strive for high recall by showing a diverse array of retrieved items rather than many similar ones–exploring how to resolve this tension is an interesting area for future work.”

In sum, the above findings and recommendations explain why platforms such as RapportiveSeriously Rapid Source Review (SRSR) and CrisisTracker add so much value to the process of assessing the credibility of tweets in near real-time. For related research: Predicting the Credibility of Disaster Tweets Automatically and: Automatically Ranking the Credibility of Tweets During Major Events.

Automatically Ranking the Credibility of Tweets During Major Events

In their study, “Credibility Ranking of Tweets during High Impact Events,” authors Aditi Gupta and Ponnurangam Kumaraguru “analyzed the credibility of information in tweets corresponding to fourteen high impact news events of 2011 around the globe.” According to their analysis, “30% of total tweets  about an event contained situational information about the event while 14% was spam.” In addition, about 17% of total tweets contained situational awareness information that was credible.

Workflow

The study analyzed over 35 million tweets posted by ~8 million users based on current trending topics. From this data, the authors identified 14 major events reflected in the tweets. These included the UK riots, Libya crisis, Virginia earthquake and Hurricane Irene, for example.

“Using regression analysis, we identi ed the important content and sourced based features, which can predict the credibility of information in a tweet. Prominent content based features were number of unique characters, swear words, pronouns, and emoticons in a tweet, and user based features like the number of followers and length of username. We adopted a supervised machine learning and relevance feedback approach using the above features, to rank tweets according to their credibility score. The performance of our ranking algorithm signi cantly enhanced when we applied re-ranking strategy. Results show that extraction of credible information from Twitter can be automated with high confi dence.”

The paper is available here (PDF). For more applied research on “information forensics,” please see this link.

See also:

  • Analyzing Fake Content on Twitter During Boston Bombings [link]
  • Predicting the Credibility of Disaster Tweets Automatically [link]
  • Auto-Identifying Fake Images on Twitter During Disasters [link]
  • How to Verify Crowdsourced Information from Social Media [link]
  • Crowdsourcing Critical Thinking to Verify Social Media [link]

Predicting the Credibility of Disaster Tweets Automatically

“Predicting Information Credibility in Time-Sensitive Social Media” is one of this year’s most interesting and important studies on “information forensics”. The analysis, co-authored by my QCRI colleague ChaTo Castello, will be published in Internet Research and should be required reading for anyone interested in the role of social media for emergency management and humanitarian response. The authors study disaster tweets and find that there are measurable differences in the way they propagate. They show that “these differences are related to the news-worthiness and credibility of the information conveyed,” a finding that en-abled them to develop an automatic and remarkably accurate way to identify credible information on Twitter.

The new study builds on this previous research, which analyzed the veracity of tweets during a major disaster. The research found “a correlation between how information propagates and the credibility that is given by the social network to it. Indeed, the reflection of real-time events on social media reveals propagation patterns that surprisingly has less variability the greater a news value is.” The graphs below depict this information propagation behavior during the 2010 Chile Earthquake.

The graphs depict the re-tweet activity during the first hours following earth-quake. Grey edges depict past retweets. Some of the re-tweet graphs reveal interesting patterns even within 30-minutes of the quake. “In some cases tweet propagation takes the form of a tree. This is the case of direct quoting of infor-mation. In other cases the propagation graph presents cycles, which indicates that the information is being commented and replied, as well as passed on.” When studying false rumor propagation, the analysis reveals that “false rumors tend to be questioned much more than confirmed truths [...].”

Building on these insights, the authors studied over 200,000 disaster tweets and identified 16 features that best separate credible and non-credible tweets. For example, users who spread credible tweets tend to have more followers. In addition, “credible tweets tend to include references to URLs which are included on the top-10,000 most visited domains on the Web. In general, credible tweets tend to include more URLs, and are longer than non credible tweets.” Further-more, credible tweets also tend to express negative feelings whilst non-credible tweets concentrate more on positive sentiments. Finally, question- and exclama-tion-marks tend to be associated with non-credible tweets, as are tweets that use first and third person pronouns. All 16 features are listed below.

• Average number of tweets posted by authors of the tweets on the topic in past.
• Average number of followees of authors posting these tweets.
•  Fraction of tweets having a positive sentiment.
•  Fraction of tweets having a negative sentiment.
•  Fraction of tweets containing a URL that contain most frequent URL.
•  Fraction of tweets containing a URL.
•  Fraction of URLs pointing to a domain among top 10,000 most visited ones.
•  Fraction of tweets containing a user mention.
•  Average length of the tweets.
•  Fraction of tweets containing a question mark.
•  Fraction of tweets containing an exclamation mark.
•  Fraction of tweets containing a question or an exclamation mark.
•  Fraction of tweets containing a “smiling” emoticons.
•  Fraction of tweets containing a first-person pronoun.
•  Fraction of tweets containing a third-person pronoun.
•  Maximum depth of the propagation trees.

Using natural language processing (NLP) and machine learning (ML), the authors used the insights above to develop an automatic classifier for finding credible English-language tweets. This classifier had a 86% AUC. This measure, which ranges from 0 to 1, captures the classifier’s predictive quality. When applied to Spanish-language tweets, the classifier’s AUC was still relatively high at 82%, which demonstrates the robustness of the approach.

Interested in learning more about “information forensics”? See this link and the articles below: