Abstract
Throughout the COVID-19 pandemic, a plethora of false information has flooded social media and online platforms, causing widespread panic and disarray among surrounding communities. Such posts contain misleading pseudo- knowledge, which typically aim to coax users into dismissing the severity of the virus. As a result, many individuals have decided to ignore important safety protocols, such as wearing masks, maintaining proper social distancing, or receiving the recently developed vaccine. This paper offers a potential solution to slowing the spread of this viral misinformation. The solution is a factual analysis computer program that accumulates reliable knowledge from previous experiences of flagging false information. The program will operate on the Python programming language and will have the ability to identify misinformation based on key phrases made by users. Such phrases will be stored in a continuously growing library within the program, then accessed when a post must be flagged. This will hopefully bring some order to the chaos caused by social media, which will, in turn, provide users with honest content that encourages decent public health practices.
Keywords: Misinformation, Social Media, Python, Technology, Artificial Intelligence, COVID-19, Software, Fact-Checking, Program.
1. Introduction
At the beginning of the coronavirus outbreak, much was uncertain. That is, there wasn’t a lot of reliable information being passed around regarding the nature of the virus. Consequently, new symptoms were popping up left and right, safety protocols were constantly changing, and society was left in total confusion. Unfortunately, in this day and age, with confusion comes absolute mayhem, particularly across social media. People began developing conspiracies against the virus and posting them on the internet, where their nonsense could attract tens of thousands of views within just a couple hours. What’s worse is that viewers could share these posts for even more people to see, gaining more and more attention until it became the headline of a local news station. This gossip was not only fictitious; it was also a massive threat to public health. All kinds of posts were being made; there was talk of COVID-19 being a ploy by the government to instill fear into citizens and keep them in line. As a result, some people started to ignore guidelines established by the Center of Disease Control (CDC), like wearing a mask or maintaining proper social distancing. Others began to feel helpless and started stockpiling household supplies until local grocery stores were wiped clean. When a vaccine was finally developed, rumors of government tracking chips found within the vaccine began making their rounds as well, which caused a significant decrease in the amount of vaccines administered.
The problem here is that most social platforms offer no means of controlling this misinformation. Users are subject to no form of accountability or factual verification, and as a consequence, user integrity is always up in the air. Obviously, there is a dire need for some kind of damage control; thus, this paper aims to create a program that auto-assesses post credibility by detecting key phrases used within a post. These phrases will come from a library coded within the program, which will be accessed when a post must be flagged. A flagged post will signal the use of misinformation and will cause the software to post a reply, stating that its content is false, then report the user for spam. Additionally, the program will be able to continuously update its library though experience until it becomes the perfect fact- checking machine. With this reliable software in place, social media will become a safer place for people to properly educate themselves on honest information regarding COVID-19. Not only will this lessen ignorance, but it may bring a quicker end to the worldwide coronavirus pandemic as well.
2. Related Works
This is not the first attempt at controlling viral misinformation. In 2016, The Association for Computing Machinery (ACM) attempted to train machine-learning (ML) algorithms to find rumor-correcting tweets as well as identify the rumor it was correcting. As their research states, the purpose of this software was to “leverage the ‘self-correcting’ crowd” [1]. Essentially, what this means is that they were trying to model their software after the honest members of the Twitter community, who were quick to point out bogus tweets. This proved to be ineffective, however, since these members seemed to be so few and far between. In addition, the program was slow and ended up lagging behind the spread of the misinformation.
The ACM also talks about formal crowdsourcing, which “distribute[s] messages to paid or volunteer crowdworkers and ask[s] them about each post’s credibility” [1]. In theory, this sounds like a pretty decent solution; unfortunately, though, this system could easily be manipulated if a crowdworker is pressured by extenuating circumstances like bribery. Additionally, ACM mentions using crowdworkers to calibrate the aforementioned ML algorithms, but this may result in the same kind of manipulation as the previous method.
The majority of this paper’s research, however, builds upon an ML algorithm developed by the ACM that selects specific features from a post and runs them through a credibility check, comparing them to factual data found within the program. Although, the software will occasionally flag a legitimate resource, it serves as a good starting point for developing a solution to the issue at hand.
3. Methodology
A list of specific methodologies, shown below, have been used to make this software. When a post is evaluated, the program will immediately search its contents for key phrases that signal misinformation. If the post contains a match, the program will send a reply with the phrase “this post has been flagged for false information. Please ignore its contents,” and then report the user for spam.
3.1. Detecting Misinformation
The first step is to determine whether or not the post in question contains false information about COVID-19. This is the most crucial part of the system. To construct the program, I have used Python, since it is the programming language I am most familiar with. By utilizing ‘If’ statements, we can run the post’s text through user input and search it for key phrases. The user may feel free to add to the library of phrases at any time. Figure 1 shows a sample of the code for detecting misinformation.
Figure 1: Sample Code for Detecting Misinformation
3.2. Sending a Reply
In order to post a reply, the program must have full control of the mouse and keyboard. Luckily, certain modules have been created for these very functions. The “mouse” function, created by GitHub user boppreh, allows the software to simulate mouse movements and clicks [2]. By utilizing this library, the program will be able to move the cursor over the text box, click inside, and send the reply. As for the keyboard, we will be using the “pyperclip” function, created by Al Sweigart [4]. This library will give us the ability to copy the reply from the program and paste it into the text box. Follow Figure 2 to see how the program might post a reply via Twitter.
Figure 2: Sample Code for Sending a Reply via Twitter [2] [4]
3.3. Reporting a User
The last step is fairly simple. One of the modules used for sending a reply via Twitter will also be used to report the user for spam, which is the “mouse” function. This module will move the cursor over the user options menu, click the drop- down, then select the following: ‘report tweet’, ‘it’s suspicious or spam,’ and ‘the account tweeting this is fake.’ See Figure 3 for the sample code.
Figure 3: Sample Code for Reporting a User via Twitter [2] [3]
4. Model Design
The design of the software is fairly simple, although, its use may require some coding knowledge. Since the program runs on Python, the user must have Python installed on their computer. Obviously, the user must know how to start the program. Additionally, when a post is being evaluated, the user must also know how to enter the post’s contents into the software as well as add to the library of key phrases. The user may also require a supplementary resource to find the cursor’s position on screen, as this is crucial to using the “mouse” module. Personally, I would recommend using a program developed by eFigureOut called Mouse Locator [3]. This software should suffice in finding the correct X and Y coordinates required for the code, which the user must also know how to enter.
The software is made to be user-friendly. This can be seen upon startup, where the user is greeted with a welcoming message, then prompted to enter their post. After the post has been pasted into the program, it is checked for key phrases that signal misinformation. If any are detected, the program will flag the post, send a reply, and report the user accordingly. If none are detected, the program will end (Figure 4).
Figure 4: Model Design
5. Controller Design and Implementation
The software was built using Python and utilizes two libraries to help it run efficiently. The system works in steps; the user must begin by entering the post and all of its relevant coordinates into the program. The next step is to run the code. Running the code can result in two outcomes. The first outcome occurs if the software detects misinformation, where it will promptly send a reply to the post’s author, then report the author for spam. The second will occur if no misinformation is detected, which will cause the program to simply end. Figure 5 shows a flowchart of controller design.
Figure 5: Flowchart of Overall Controller Design
6. Cost
There is no additional cost to use this software; it is free for anyone to run at any time. Also, any person may modify the program according to their personal preferences as well as upload their modifications for the use of others.
7. Conclusion
This paper finds that the software can identify misinformation on social media regarding COVID-19. It discusses the functional ability of the program where it will auto-assess post credibility by detecting key phrases within the text that signal the use of false information. Additionally, it will send a reply, stating that the post is unreliable, then report the user for spam. The paper discusses in detail how misinformation is detected, how a reply is sent, and how the user is reported. Related works are also explored, which served as a basis for developing the program. The paper ensures that the software is cost-effective and open for modification.
Acknowledgments
The author would like to thank Mr. David Johnson from the English Department at East Carolina University.
Conflicts of Interest
The author declares no conflicts of interest regarding the publication of this paper.
References
[1] Association for Computing Machinery. (2016). Slowing the spread of viral misinformation: Can crowdsourcing help? The Huffington Post.
[2] Slowing The Spread Of Viral Misinformation: Can Crowdsourcing Help? | HuffPost (2020). mouse v0.7.1. GitHub.
[3] GitHub – boppreh/mouse: Hook and simulate globalmouse events in pure Python (2016). Mouse Locator v1.0.0.1. Softpedia. https://www.softpedia.com/get/Others/Miscellaneous/Mouse-Locator.shtml
[4] Sweigart, (2021). pyperclip v1.8.2. PyPI. pyperclip · PyPI
write