AI Moderator Tools Free Essay Example

Some of the things that AI Moderator Tools miss when they’re Applied to User Generated Content.

The current argument for AI moderation of user generated content seems to be one partly driven by the big social media platforms as they try to keep up with subversion of their terms and conditions and sinister activities such as terrorism, racist agendas and the spreading of misinformation and “fake news”. But it’s important at this point to look to find some perspective.

The decisions being made by massive platforms such as Facebook are based on a need to moderate what are almost indescribably big volumes of user generated content, made up of multiple forms of media – video, pictures and memes, and text in a multitude of different languages.

It’s the sheer volume of this task that has forced the hand of people like Mark Zuckerberg to employ AI, and it’s by no means an off-the-shelf or a perfect solution. This is a reaction based on need, and in fact, it’s nowhere near to being adequate as a means to moderate user generated content.

It can, however, be viewed as a good barometer of the technology and the abilities and limitations of machine learning based systems to moderate human behaviors and interactions.

Government pressures to censor

Some of the more recent changes being brought by the largest social media platforms are driven by government attempts to control information being shared by terrorist organizations. Such government moves could be seen to be shutting the stable door after the horse has bolted, so to speak.

In many cases, problems with social unrest are just that – social problems caused by government policies, wars and exploitation in world politics. The fact remains, however, that the spread of unrest has spilled onto social media and the viral nature of shocking images and information, is a problem for those tasked with maintaining law and order.

To a degree, from an outside observer’s point of view, these changes and the implementation of AI can easily be considered to be desperate platforms, scrambling to keep up with government demands and threatened sanctions. Whilst the big platforms may have no choice but to compromise fairness, inclusivity and accuracy in the flood of almost unsurmountable volumes of user generated content, the potential pitfalls for those that host locations for discussion, customer input and feedback, are those of alienation of customers, potential clients, and even ethnic and religious groups. The temptation to employ AI where volume doesn’t demand it should be avoided, because what Facebook has taught us is that AI isn’t ready – and that it may indeed never be ready, to deal with the complex challenges involved in interpreting the way in which humans interact, debate, and pass comment.

One of the biggest indicators when viewing Facebook’s recent efforts to employ AI from afar, is their subsequent need to recruit thousands of human moderators. If we ask ourselves the question, “Who’s watching the watchers?” – then the answer is; humans. And there are very good reasons for that.

Written and spoken speech, and AI

Let’s consider for a moment the way in which humans communicate and express themselves and convey their opinions via written text and the spoken word. How does AI interpret and make decisions based on this? Machine learning is only as effective as the training it receives for a specific task. When it comes to moderating interactions between humans on a global scale, such as on the internet, the rules need to be ridiculously complex to begin with. Add to that, the problem that there are many forms and dialects within just the English language alone, and that it takes humans a lifetime to learn mainly just their own local form, and then the task ahead of the machine is an impossible one. To develop a tool that can operate across every form of communication, language, and each dialect, is the stuff of fancy.

Natural Language Processing tools are trained by humans in order to understand the type of text they’re examining, but this understanding is limited to fairly simple rule making. A trainer may, for instance, input many texts into an NLP tool that are tagged and classified by the trainer into two defined groups.

One example of this would be pro, and anti-terrorist sentiments, including statements meant to radicalize. The tool will attempt to learn over time, the pointers and flags within such texts that make it likely to belong to either of the two defined groups, and in that way, an NLP tool will develop a system for identifying pro-terrorist user generated content. One of the big limitations with this form of NLP moderation is the lack of any ability within the tool to discern the intent of the statement maker, and therefore NLP poses great scope for inaccurate censorship.

I come from a city in the UK called Liverpool. The accent and dialect is, as in a lot of places in the world, very unique. You have to be born and raised there to have any hope of ever understanding all of what is being said to you or discussed in your vicinity. Phrases, the way words are used, and their intended meanings can differ in the space of a few streets and from neighborhood to neighborhood – as can accents. The whole thing can be a minefield of communication and misunderstandings.

The UK is a relatively small place. Dialects change dramatically within the space of a few miles. Many people who visit and travel around England, are bemused and confused by the differences in the way in which people speak, and by the local meanings attributed to specific words.

The present reality of AI and NLP is that to impart any sort of understanding of these subtle, and multiple, linguistic differences to a machine would require almost inconceivably large amounts of input and training data. And you’d have to find humans to understand it all in the first place. Broad NLP tools that operate within all areas of the worldwide web, and monitor all languages and communications, are not feasible. AI would have to be more advanced, and the realm and training of each tool would have to be way more localized, specialized, and more expensive.

Another factor that generates differences in the way people from around the world communicate with each other is culture. What means one thing to me, and is considered polite and respectful in the part of the world in which I live, may be an insult in another part of the world, or to someone that practices a different religion – or indeed, no religion. We all know that the possibilities for misunderstandings between people from vastly different cultures is massive in the real world. Online, that potential for offence or fallings-out is greatly amplified by the fact that communication is instant, be it over a distance of ten miles or ten thousand, and heightened especially by the fact that communications, especially on a public forum, can take on the form of discussing culturally sensitive subject matter, and be conducted between thousands of users at once. These participants might all be from differing geographical locations and have differing belief systems and contrasting moral or civil codes. It’s a powder keg that requires very subtle moderation and great understanding and attention when things get heated and passions are high. Moderating human communications is difficult and it requires a special set of skills – but crucially, it also demands real insight and carefully considered judgement. Getting it wrong will nearly always be judged by one party as being biased, or even worse, unfair censorship. There’s the potential that bad moderation could be viewed as culturally bigoted. The implications of that are potentially very damaging to a company or an organization that provides a forum for public interaction and discussion. And those sorts of negative incidents can rub off very badly indeed.

Forums and social media sites can be great, productive cauldrons of ideas and promote better cultural understanding. They have the potential to be real tools for building a better world and improving the way in which we all understand each other. But that great, ongoing potential brings with it the need for great moderation. And effective progress in the way we understand other cultures and share ideas can only happen if the users are largely all there with the right intentions. Humans often find it difficult to find a troll. Especially when emotions run high. Solutions to problems are only found when problems are discussed, and sincerity and genuine attitudes combine. It’s sometimes very difficult to figure out who has sinister motives and might be sprinkling too much salt in the melting pot. Likewise, nobody wants a hand of moderation so heavy that it acts on too wide a set of rules and parameters and excludes those with genuine intentions of good discussion and ideas. That is not how progress occurs.

Sarcasm

It’s an everyday part of the way humans interact, and it’s usually pretty apparent to us when somebody uses sarcasm to make a point. For machines, though, the process of identifying and interpreting the intent of sarcasm is a very complicated business indeed. AIs rely on masses of data input when applying NLP to the task of moderating UGC. The rules laid down for the process and the data provided by human programmers is finite and often hard to source. Let’s look at a statement such as the following:

“The universal TV remote I just bought from Acme Remotes is amazing! It’s so universal that when I attempt to change the channels on my TV, it shuts the blinds.”

Now, me and you might be able to instantly identify obvious sarcasm in that sentence, but what would an AI employing NLP to monitor UGC in a comments section make of it?

NLP works by breaking down the components of a sentence and then analyzing the meanings of words within it. Sarcasm is an area of human expression that, by definition, doesn’t conform to the practice of using literal meaning. It creates a micro-context that is foreign to the way in which we communicate conventionally, yet in its own right, is a very effective form of communication. It’s usually missed or misunderstood by AIs.

Pictures and Memes

Pictures and Memes are a difficult area for AI. Human moderators look at an image, or an image that incorporates some form of text, and they pretty instantly comprehend all of the components that make it up.

Memes are slightly more context-reliant and subtle forms of visual communication that often involve a generous portion of satire and/or sarcasm to make a point, or a joke. Whilst Memes can be powerful tools for both good and subversive intents, they often take even the more seasoned human moderator some study time to work out and to assess, in terms of suitability for a particular online environment.

If we consider the example of a picture of a protest in which a protester holds up a sign. Such signs can often be fairly ambiguous when taken out of a known context. An image such as this may appear in a news article about let’s say, an anti-Semitism protest, and a human reader or moderator will instantly recognize an offensive phrase or word on such a protest placard to be a declaration in support of Jews. An AI though, relying on a combination of programming and a series of techniques to first break down the image into components of text and picture elements, and then assessing each element separately, without reference to context, might completely miss the point of the message on the sign. The challenges facing AI developers in terms of producing technology that can operate across and within several subtle and intertwined layers of human communication and behaviors, are huge ones.

Lies and Fake News

Fake news. It’s a real modern-day preoccupation. With the rise of social media platforms and a multitude of online news outlets, misinformation has arguably become more prevalent. Propaganda isn’t a concept that was conceived any time recently however, and rather than an increase in instances of misinformation, what we’re probably observing is an increase that’s merely proportional to the volume of news and information that the internet has made available to us. It’s one of the great indicators of an increasing need for effective moderation, though. And all indications so far seem to illustrate that it’s a battle that AI alone isn’t capable of winning.

Beyond overly reactive forms of machine-based censorship and blanket banning of particular news outlets and monitors on geographical areas known to produce fake news, the task for AI based moderation of such news content is a very difficult one.

There have been schemes set up whereby AI filters content for the purpose of passing suspect articles to journalists for review, but this amounts to little more than a glorified – and expensive filter. It’s the humans that we rely on ultimately, to discern correctly what’s true and what’s bogus.

It’s probably telling that, in an area of moderation where the heads of presidents can roll on the back of the wrong piece of fake news, or at least on the basis of a well-organized campaign of misinformation and dubious news stories that whip up public opinion, AI has been limited to the role of assisting human moderators.

Bias

Depending on how you consider the use of technology and computers, this is a weakness in AI moderation that may surprise you. However, it’s not as surprising when you remember that AIs rely on data input from humans. And as when we ‘programme’ our own offspring, bias gets passed on to some degree or another.

Now, I’m not suggesting that AI developers are a screaming bunch of racist, xenophobic terrorist lunatics, but it’s unavoidable that some of their preconceived ideas and ideologies get passed to the models.

AI teaches itself, based on the data it is supplied with. There’s even reason to believe that minority groups are worse off under a system of NLP moderation, because even given a broad and representative range of input data, there’s proportionately less data about minority groups, and therefore less potential for the AI to develop an understanding of them, which can result effectively in a form of bias.

The success rates we can currently expect from NLP based AI are so weak, that in order for a rate that’s acceptable on social media to be attained, filters need to be set so high as to apply bias, and in the relative absence of any bias, what may still be perceived to be biased to a degree that alienates minority groups. The technology just isn’t advanced enough to be accurate and maintain universal fairness.

Caution is required

There’s no doubt that webmasters, business owners and those responsible for upholding order on websites, forums, comment sections and social media need to be careful when choosing how they’ll moderate user generated content. There’s a temptation to view the volume of work that the AIs, and in particular, NLP can get through – quickly. However, whilst that attraction is somewhat understandable in some ways where massive volume exists, there’s very good reasons to consider the inaccuracies that are undeniable with such systems and view the results of that as extremely negative.

Unfair moderation, irrespective of the reasons for that unfairness, can quickly alienate customers and users to the point of causing a problem. Use of AI to moderate user generated content should be limited by its failings. And if we apply that rule, then it’s use needs to be limited extensively, because the failings of AI in such areas, are in themselves extensive.