Pose governance on AI algorithms in social media platforms

Main Theme

This blog briefly outlines an overview of AI algorithms application in social platforms and highlights the necessity to strengthen governance. Take AI-swap technology as an example, and see how AI technologies were applied and generated Deepfakes. By scrutinizing AI’s legitimacy in social media platforms, this discussion seeks to shed light on the potential directions in the future.

Intro: The era of AI

In recent years, the number of users of social platforms has increased rapidly, with data reaching more than 4 billion. Recent statistics indicate that 4.95 billion people now use social media worldwide, up more than double compared with data from 2.07 billion in 2015 (Brian Dean, 2024). The proliferation of social media is inseparable from the support of Artificial intelligence (AI) algorithm technology. Indeed, AI algorithms and social media are a perfect match and this combination has greatly promoted the prosperity and development of the Internet ecosystem.

Frankly speaking, almost every dominant social media platform in the Internet ecosystem now uses AI in some way to function as a support to its operation. This is all to say that AI is becoming a fundamental part of today’s social network’s function. However, as the user number is still surging, the inherent shortcomings of AI are becoming increasingly apparent, attracting people’s attention worldwide. Unlike traditional mass-media, algorithms are designed on a global level embedded in most popular websites worldwide (Natascha Just, Michael Latzer, 2017). Currently, more and more people are concerned with the negative impacts of AI algorithms within social media platforms such as user data leakage, information cocoons, public opinion manipulation, etc.

Becoming a Daily Necessity: social media

According to the statistics, the average time spent on social media platforms per day is approximately 2 hours and 24 minutes worldwide among users aged from 16 to 64, across various devices (Brian Dean, 2024). Using social media apps is not only enjoying a popular lifestyle but also becoming a Daily Necessity embedded into people’s lives.

For this blog, I prefer defining the concept of ‘social media’ as an Internet-based form of communication enabling users to start conversations, share information, and create content. There are many different forms of social media, including blogs, micro-blogs, social networking sites, photo-sharing sites, instant messaging, video-sharing sites, podcasts, etc. The mainstream social media platforms involve many different fields, for example, Tik Tok for short video content; Instagram for photo sharing; Facebook for social connections; and X (formerly known as Twitter) for mini-blogs, each designed to cater to a particular user’s needs.

Compared to the static web portals during the Web 2.0 era, nowadays social media platforms show their unique characteristics in common and provide broad space for AI algorithm applications. According to Mayfield’s view, she had identified five specific characteristics that underline the operations of all social media: participation, openness, conversation, community, and connectedness (Mayfield, 2008).  As social media continues to evolve, especially integrated with cutting-edge technologies, the convergence of AI algorithms has become a new power promoting platform development in current.

A match made in heaven: AI algorithm and social media platform

Clearly, AI can vastly improve the social media experience for users.  That’s mainly due to that AI can help practitioners increase productivity and performance in the workflow of platforms.

As defined by Investopedia, Artificial intelligence (abbr.AI) refers to the process of simulation in which human intelligence was applied in machines programmed in advance to act like human beings (The Investopedia Team, 2023). Through pre-set algorithms, computers can recognize different patterns and images, receive task instructions, make decisions, and exhibit judgments just like humans. And this is what we distinguished as ‘intelligence’ in common.

In the realm of social media, practitioners mainly use algorithms to analyze user behavior, preferences, and feedback. With data collected from users, AI can make better predictions and adjust to user’s needs, optimizing operation strategies simultaneously. With that in mind, let’s dig deeper into how AI is applied to social media platforms. Surbhi‘s report summarizes 10 ways to apply AI in social media, improving the experience for both operators and users (Surbhi in Sprinklr, 2023).

More broadly, let’s introduce the concept of workflow to understand the application of AI in a better way. As defined by Chun Ouyang et al., workflow denotes a series of tasks (activities) through which work is routed (Chun Ouyang et al.2010), usually necessary to achieve a given goal. In brief, it refers to the process of completing tasks, as for social media, namely equals basic operation.

Analyze the workflow as shown in chart 1, obviously, AI can be applied in every step of the process. At the first stage, namely the content creation stage, AI algorithms can create blog posts, captions, and other content in batches during a set period of time. Next, when dealing with content processing, it mainly relies on recommendation systems. This is the most common and widespread application case. Then, AI-generated content can be disseminated to users on different devices through multiple channels. Finally, AI can make a comprehensive analysis based on collected user data through the whole workflow.

In summary, the AI algorithm has great application capabilities at every stage of the social media workflow. Probably that is the reason why we think that AI algorithms are everywhere. AI and social media complement each other, and the entire Internet ecosystem is developing rapidly. Although technology is neutral objectively, there are still some technical drawbacks emerging in its application.

Case study: Unavoidable Deepfakes

• Assumption is becoming a reality

First let’s make such an assumption: one night you feel bored and lonely, then you log in to a social media app receiving a private message that “I want to be your new friend, send you some photos (usually naked)?” from a female account with an avatar looks like a beautiful girl with scantily clad.

What will you do in this situation? Continue or reject?

Probably most of you will choose to continue this conversation. But what if the “new friend” is an AI-generated account? What if the dizzy photos are all technologically synthesized? Have you ever thought about the possibility that your new friend is totally fake and even never existed?

In fact, these behaviors are quite pervasive now. The development of science and technology has brought much convenience to people’s lives, and also some absurd possibilities. There are two forms of the “fake friend”: The first one is to steal your personal information and using AI to generate extra content, creating a duplicated version of you. Another is to train AI with images and then generate a virtuous person who never existed. Both of them can be defined as Deepfakes.

• AI face-swap technology

One of the most interesting examples in the application of AI algorithms is AI face swapper technology which allows you to swap faces in seconds effortlessly with the preset photos. As the name suggests, use AI you can change the faces of the particular photo with other people’s faces. For example, with AI technology you can change Emma Watson’s face into a photo of yourself and create a new poster without much effort.

Does it sound interesting and appealing? And it is easily doable.  Moreover, the process is low-cost, low-barrier, and also high-efficient. If you search keywords like “face change” or “face swap” on the Internet, many websites and apps providing such a service. Even a primary student can use AI to generate Deepfakes with other people’s faces.

• Deepfakes threatens

Although there is a beneficial side of AI algorithms, the use of this technology still brings a lot of data security issues and social controversy. Let’s think on the other side: what if you applied face-swap technology to pornographic images and videos without your permission? What would you think when “new contents” were created and sent to your friends?  Do you still think this technology is interesting? Sadly speaking, this problem has become widespread and many people are suffering from it.

Based on the research company Sensity AI estimates, between 90% and 95% of all online deepfake videos are nonconsensual porn, and around 90% of those feature women (Karen Hao, 2021).

According to my observation, there are already many self-created porn platforms on the market that use AI face-swap technology. For instance, a website called “Reface Porn” acclaims that you can do face change, face swap porn and make deepfake porn videos online with their app.

According to the Weibo content of CCTV News, some people use AI face-swap software to generate a number of pornographies and spread them all over the Internet, thus gaining profits from it. After this kind of harmful incident occurred, many users could not help but start to fear the application of AI technology. Setting restrictions on the application of AI technology in social media platforms has also become a topic of great concern.

AI algorithms: a double-edged sword

Despite all these advantages, AI algorithms are somehow perceived as a “black box” where the instructions made by the algorithms are sometimes biased and lack transparency, even perpetuating racism and sexism (Noble, Safiya Umoja, 2018). Undoubtedly, AI algorithms can be regarded as a kind of “double-edged sword” technology when applied to social media platforms. As Karl Jaspers concluded in his philosophy of technology, technology is inherently neutral in itself, no more than a means for human goals, for it is incapable of generating its own goals. This neutrality makes human beings responsible for what they make of technology (Verbeek, Peter-Paul, 2005). There are both advantages and disadvantages of AI algorithms, depending on the application method and particular social media context. We can go further for the specific pros and cons.

On the positive side, the utilization of AI algorithms in social media platforms could improve the overall operating efficiency of the particular app. In traditional mode, the practitioners spent a lot of time checking errors and debugging. Now they only need to preset code and dispatch instructions. In addition, applying the automation of AI algorithms saves costs significantly. As a consequence, the AI algorithm has become a core technology part for Internet companies due to its significance in supporting the basic operation of the platforms.

More and more companies are increasing their investment in research and development of AI technology.  For instance, Meta and IBM are co-launching a new ‘AI Alliance’ that brings together AI-curious companies across a range of industries. Amazon launched an AI platform aimed at corporate customers who want to incorporate AI into their businesses. Facebook has configured an AI chatbot as an embedded service accessible to users.

On the other side, some potential drawbacks of AI algorithms cannot be underestimated similarly. The most concerning issue refers to user data security because it is relevant to everyone who uses social media. That is to say, AI cyber security risks should be treated as critical. At the same time, the popularity of social media among teenagers (usually range 13 to 18) cannot be underestimated similarly. Therefore, the information cocoon phenomenon is a very prominent issue in the use of social media platforms. According to Sunstein‘s view, information cocoons create an environment in which people only encounter voices that express opinions and ideas similar to their own. We may also use an “echo chamber” which stands for a similar problem in that content is repeated and reinforced to describe the same situation (Chen CF, Shi W. 2018). Moreover, there is another problem that should not be ignored — AI ethics which refers to privacy and surveillance, bias, and discrimination. Objective AI ethics in technological applications can help to build a world with less bias and more fairness.

The necessity of AI governance

Based on the factors listed above, governance for the application of AI on social platforms has become an inevitable trend. Next, I will state the necessity for AI governance from different perspectives.

For users surfing social media platforms, applying governance on AI algorithms means strengthening data security and providing a safer cyberspace where we could enjoy the convenience that new technologies brought to our life.

For platforms, governance means a way of external management and it can be a great help for fostering the development of platforms themselves in a rational and healthy path.  Lack of governance will lead to a situation in which users may easily lose trust in the platforms and thus cause user loss.

For the technology itself, only by using AI algorithms in the right way can technology not get out of control. The nature of technology is often affected by its application. The key to evaluation depends on how we use and how we control it.

For humanitarianism, governance of AI technology is an important measure to prevent technology from getting out of control. With a humanism philosophy, we should never let technology take the lead and better be rational when making decisions, especially concerning social media platforms where it has a wide scope and far-reaching impacts.

Possible directions for AI governance

Based on my personal understanding, I think governance on AI algorithms can be considered from the following directions:

In conclusion, the governance of AI is still being explored, requiring efforts from multi-stakeholders to reach a census. The use of technology is closely related to each of our lives in the context of digital media and still has a long way to go. 

Reference

1. Brian Dean, Social Media Usage & Growth Statistics, https://backlinko.com/social-media-users#how-many-people-use-social-media, 2024
2. Natascha Just, Michael Latzer, 2017. Governance by algorithms: reality construction by algorithmic selection on the Internet, 2017, pp. 250 Media, Culture & Society 2017, Vol. 39(2) 238–258
3. Mayfield, A. (2008). What is social media? iCrossing. Retrieved August 25, 2009, from www.icrossing.co.uk/…/What_is_Social_Media_iCrossing_ebook.pdf
4. The Investopedia Team, Artificial Intelligence (AI): What It Is and How It Is Used, from https://www.investopedia.com/terms/a/artificial-intelligence-ai.asp#:~:text=Artificial%20intelligence%20(AI)%20refers%20to,as%20learning%20and%20problem%2Dsolving,2023
5. Chun Ouyang, Michael Adams, Moe Thandar Wynn, and Arthur H.M. ter Hofstede, Workflow Management, 2010
6. Noble, Safiya Umoja. Algorithms of Oppression : How Search Engines Reinforce Racism, New York University Press, 2018, pp.29 ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/usyd/detail.action?docID=4834260.
7. Verbeek, Peter-Paul. “1 Technology and the Self”. What Things Do: Philosophical Reflections on Technology, Agency, and Design, University Park, USA: Penn State University Press, 2005, pp. 15-46. https://doi.org/10.1515/9780271033228-003
8. Chen CF, Shi W. Technical interpretation and value discussion of personalized news recommendation algorithm. Chin Edit J. 2018;10:9–14.
9. Karen Hao, A horrifying new AI app swaps women into porn videos with a click, MIT Technology Review, from https://www.technologyreview.com/2021/09/13/1035449/ai-deepfake-app-face-swaps-women-into-porn/ ,2021