AI in Social Media Marketing: From Social Bots to Object Recognition

As we have shown in the previous parts of this series, Artificial Intelligence is becoming increasingly important in online marketing. It is no surprise that the social media world is also influenced by it. Besides Microsoft's failed chatbot experiment, there are numerous other applications where AI is already part of everyday life. Often unnoticed by users, AI-based data evaluation and personalized content distribution play an important role.

But what about the production of content and communication? Can machines replace humans? We take a closer look at three selected platforms, their use cases for artificial intelligence, and the developments we can expect in the future.

Facebook

For years, Facebook has been investing in AI research. Currently, "deep learning" is used, for example, to personalize displayed ads and suggested pages as well as to generate the personal news feed. AI also plays a role in the further development of the Messenger as a tool for customer communication. The future could include automated responses to messages, sending relevant information, products, or images depending on the inquiry. In image recognition, Artificial Intelligence ensures that Facebook announced in 2014 an ability to correctly match faces with an accuracy rate of over 97%. Through manual tagging by users, the system continually learns and achieves this remarkably high accuracy.

Pinterest

Thanks to its partnership with the startup Kosei since 2015, Pinterest has also improved the interpretation of images using machine learning. Automatic object recognition allows the platform to deliver personalized pins and product suggestions based on relevance calculations. These calculations consider not only a user's clicks and repins but also language and location (country). At the beginning of this year, the company announced that it would revise the Related Pins feature using deep learning to achieve even more relevant results and higher user engagement. Object recognition is also intended to enable users to identify objects, such as products, in images—and buy them. To further advance developments in machine learning and AI, Pinterest Labs was launched this year. With the help of renowned researchers and scientists, Pinterest developed an algorithm responsible not only for the order of pins and search results but also for placing relevant ads, filtering out spam content, and learning to understand the content of pins.

LinkedIn

Since Microsoft purchased LinkedIn last year, AI is likely to gain even more importance for this business network. Since 2014, machine learning with the Bright technology has been used to show employers suitable candidates and generate relevant job suggestions for job seekers (and freelancers). Again, success is determined by relevance (based on a variety of criteria, such as location, previous jobs, and profile keywords). A self-learning algorithm can continuously improve matching based on reactions to suggestions, offering both employers and employees the best possible relevance—and increasing the chances of finding a suitable job or candidate. Unless Xing keeps up, LinkedIn clearly has the advantage as a portal for job and employee searches.

Social Bots

The advantage that Artificial Intelligence offers social media platforms (and thanks to higher relevance of content and recommendations also to users) is evident in the three selected examples. But how can companies use AI in social media marketing? What automation opportunities are available? Similar to content marketing, the question arises: Can an intelligent system independently manage social media accounts, produce and distribute content? Will social media managers become redundant as a result?

Microsoft's example, which made headlines last year, gives reason for skepticism, at least at this time. Microsoft's chatbot, created to represent the artificial teenage girl Tay, had to be shut down after just 24 hours because it made racist and offensive remarks. What happened?

The system learned from communication with other users, who took the opportunity to teach "Tay" suggestive and other morally questionable language and statements.

Ultimately, machines can only be as "moral" as the data foundation allows. Should AI-based systems ever autonomously manage accounts, very controlled training, the definition of specific keywords, and guidelines regarding tone and "opinion" are required to ensure that corporate principles are accurately represented and general rules of human interaction are not violated. To what extent or when companies will take such a risk to fully automate social network activities is questionable.

Social Analytics

A significantly less risky but still not uncritical application of Artificial Intelligence is in the field of social analytics. Whenever it comes to analyzing large volumes of data, AI's great strength comes into play. A promising possibility is to analyze data related to specific topics, products, or companies and derive actionable recommendations from it.

However, even this is problematic at the current time: Evaluating status updates for positive or negative sentiment is prone to error since, for example, not everyone underscores their Facebook post with an emotional status code. Humor, irony, or cultural differences can make assessment difficult. Differentiations are also difficult to automate. This challenge is known from media monitoring even before the advent of AI. The less human control is integrated into the process, the more error-prone the results. Therefore, manual quality control is a decisive criterion for the reliability of the evaluation.

Using AI Against Internet Trolls

In addition to the automatic detection of spam, there is another real application for Artificial Intelligence from which users and companies alike can benefit: One of the projects of Google's sister company Jigsaw deals with combating "trolls" and cyberbullying. Machine learning can help operators and moderators in social media marketing filter out abusive comments and hate, and then delete them or respond to them. In this scenario, AI does not take over communication with customers, fans, and followers, but supports social media managers in better managing their accounts.

Outlook: The Artificial Social Media Manager?

Social media, as the name suggests, thrives on the social aspect and relationships, such as between a brand and its fans. Should Artificial Intelligence independently take over the management of accounts in the future, this would mean that real people build a relationship with a machine that is based on data and the content and contexts taught to it. Besides the fact that a significant level of trust in technology is necessary to let a machine represent a brand, the question arises: Doesn't a brand's personality get lost without the human component? To what extent is an artificially generated personality possible?

Undoubtedly, artificial intelligence can assist social media managers in distributing content and evaluating its performance, aggregating user data, recognizing patterns, and deriving recommendations for actions. While real-time becomes increasingly important (customers expect prompt communication), AI-based data analysis, filtering and prioritizing inquiries, and recommending responses can save valuable resources (and time) that social media managers can instead use for strategic direction and the creative aspect of social media marketing. Thus, the machine is unlikely to make humans redundant here either, but instead, provide them with a tool to work more effectively and customer-oriented.