-14- Part of the Paradigm: A.I. Mediators
How can A.I. help us manage consensus data more effectively?
Something like a town hall discussion takes place on every large enough social media post. Right below the content, a comment section rife with opinion. We all scroll down to read, if only briefly. We might see 2 or 3, maybe 10 or more if we are motivated and intrigued enough to continue sifting through the discourse. Of course you might also see the odd degenerate post, spiteful and obnoxious, but we can ignore that for now.
The statistics show that the over 90% of users do not make it to page 2 of their Google search results. While I know this is not a direct comparison, I think it may be somewhat relevant in pointing out what the average attention span is for reading comments. To read 10 of them is probably more than average, to get to 20, 40, or 60, must be exponential leaps in rarity. With some pieces of content garnering thousands of comments, of course no sensible person will ever read all of them, not even the creator themself would have that capacity.
Is it wasted then, all those voices? Are they destined for the dustbin of internet purgatory? A digital garbage pile of words and outdated opinions? It is true that many are baseless, many are unrelated to the topics being discussed, many are degenerate shitposts, but many are also none of these things; many are thoughtful, many are kind, many are questions wanting to know more, many are evolutions on the topics being discussed, many are coming from professionals, people with expertise offering real insight, many are plainly helpful. So it does seem to me that we are wasting at least some amount of human potential with our neglect of these naturally flowing town hall discussion boards. They're everywhere, why are we not using them more effectively?
In case you have yet to notice, Amazon has included a very useful feature in the review sections of many of their product pages. They offer brief summaries of buyer sentiments on certain products with enough reviews. The summaries even include some helpful relevant key words: high quality, durable, good value, and so forth. This is truly, to my mind, a great addition for the one looking for helpful advice.
I have no inside knowledge on how the Amazon feature works, but every indication points to the use of an A.I. summary tool. That would mean an A.I. system is mediating what people are saying, taking into account the potentially hundreds of reviews and condensing them down into a bite-sized paragraph denoting the general consensus. No doubt a great strategy for the time-sensitive shopper. But I would like to see this feature evolve and spread to other platforms as well, as I am inclined to believe that it can be highly effective way to not only allow every worthwhile voice to matter in some small way, but I also believe a feature like this can disincentivize baseless provocation.
Voices that matter.
As I stated previously, thousands of comments, one reader. No one can or will ever read them all. Not only that, but we may never even scroll far enough to read more than a few of them. Depending on the nature of the comments we do happen to read, we might come away from the content thinking the majority are in general agreement, or the opposite: they all think this creator is a mad man. The point is, we can be far more effective than guesswork, and A.I. can help us achieve some valuable solutions.
What I would push for is a highly refined system to the one Amazon is currently presenting. I understand there may be certain limitations, but for this to work in full effect, I wish to assume virtually everything to be on the table.
What we need is an A.I. mediator with the power to harness not only the specific comments being posted, but one that also contains access to history and the entire internet database. If we are to use such a feature in a way that inspires us to trust in what it is telling us, then we ought to know that it is taking into account the full context of every engagement.
That would mean something like tracking user data, yes; across the platform, not across life, do not get hasty. But surely this is information that an A.I. mediator ought to account for if it is to explore a consensus in fine detail. It must tailor its own understanding of the discourse in such a way that provides us the context we would otherwise never see. I would like to know if a comment section is being inundated with users typically associated with this or that ideology, this or that community, this or that consensus. This is nothing new, people do investigations like this regularly, but they're ineffective, and cumbersome, and the results of the data always lag behind the real-time discourse. People move on from content extremely quickly, and so in order for these analytics to be of maximal use, it cannot be done by human heads, but rather needs to be accomplished by A.I. tools. Quick, efficient, instantly readable and understandable.
What is the ultimate purpose of this?
Leaving a comment on a piece of content is not the same as pressing a button to "like" it. The former is active and the latter is passive. Psychologists have studied the human tendency to follow the crowd as a way to avoid being left behind, or to avoid standing out as an outsider. If others are doing it, then it must be right, and the last thing you want is to be the person who did not read the social cues correctly, and have thus been left out of the loop. We are conditioned to be followers in that regard — very few manage to become leaders.
I do not think I am venturing far off track by saying that the "like" button is promoting a similiar type of follower mentality. A post with thousands of hearts is very likely to be better than average, maybe even brilliant, but how many of those hearts are coming from people who just thought: "Yea, everyone seems to like this, I guess I do to." It even works in the opposite direction: no hearts means there is no consensus on a things validity, therefore you might find it difficult to justify being the first to start the party.
Going back to comments for a moment, how do these differ from the passive act of heart-clicking? Crucially, there is an active element of thought that is involved in the commenting process. A comment represents a real stake at play, and forces some amount of cognition beyond our herd-following psychology. There is thought placed toward reputation, accountability, accuracy — what did I really think about that content? — other people can know your opinion, you are no longer entirely anonymous. Even a two word comment — "Great work" or "Enjoyed that" — is better than a lousy click. We developed sophisticated communication for a reason, so why are we suddenly so pleased with muting it with likes?
To tie this back to the A.I. mediator, such a tool could organize hundreds of comments, distilling them down into brief but manageable summaries of what the group consensus seems to indicate. It could also go into more nuanced detail about group associations — based on historical user data — and what those associated groups are saying and how it differs from the broader consensus. Ultimately, these descriptions can be as detailed as we wish them to be, but the point being that within the space of a few paragraphs, we can arrive at a level of context that allows us to understand a vast range of dynamic opinions.
Furthermore, the system would not be static in any way. It would work quickly and efficiently, and can be updated as an ongoing morphological entity. The summary after 50 comments will differ from the summary after 500 comments. The A.I. might even be able to speak on the follower dynamic directly: "More people seemed to agree as the comments continued to pour in." Perhaps that might also reveal some other interesting information on states of consensus and how they function socially.
I would like to address one valid rebuttal here: would this not lead to yet another example of the psychological tendency to follow the crowd? Well, perhaps to some extent yes, but I would still argue that the outcome would better serve us because the “crowd opinion”, so to speak, would become highly nuanced to go along with the A.I. mediator summary. For example, I can imagine a scenario where, as more data is being gathered, and as the summary offered becomes highly evolved and precisely laid out, any later users can interact directly with the nuances of that data, rather than the counted ticker of likes, or the most agreed upon top-comment. Generally speaking, I would expect to be able to learn more about what has already been discussed, and who has contributed to that discussion, over the course of 5000 comments, and then, if I so please, be able to comment something tailored to that overall discussion.
What I would like to stress is that when it comes to internet content specifically, speech is better than silence, comments are better than hearts. We should want people to talk to each other, not pretend to talk, or refuse to talk, or substitute the word for the click. We should want ideas, and we should look for more ways to cultivate and promote them. Ideas come through words, not through likes, and, whether you believe it or not (I happen to), there are ideas to be found in comment sections alone that are probably not being savoured — at the very least they are not being managed effectively.
The baseless provocation problem
I would be remise if I did not say something about the shitposters. Thankfully, most idiocy is easy to spot on the basis of its idiocy. We can trust that A.I. could also spot this, and devalue the opinions of the idiots based on its own observations. However, others are more clever with their antagonization of content creators. They still amount to being bad actors — trolls, deceivers, liars, maligners, anti-Creators — but I think the systems we build will only get better at detecting bad information, and reporting back on it. As far as free-speech is concerned, the solution should never be a matter of censorship or misrepresentation of views, good or bad. But a sophisticated enough A.I. mediator would be able to deliver a contextual understanding of the negativity surrounding a piece of content based entirely on history and search data — the historical database of the user in question ("your reputation precedes you"), and an internet-wide search result magnifying the credence of the claims being spouted. In short, I think we can come closer and closer to solving this problem, again with the help of A.I. and its power of instantaneous data collection.
I hope this has been an interesting topic of discussion for anyone with the capacity to have read this far in. I am a curious person analyzing the curious paradigms of this digital age, and I would like us to find better ways of fostering lucrative debate within the halls of our digital spaces. We owe it to future generations to put systems in place that will help lead us toward the best possible outcomes. Perhaps I am utterly wrong in my assumptions, but that is ultimately okay, because people will voice their own opinions on the matter — even a like or non-like is some kind of indicator I suppose — and that is how good or bad ideas are magnified or drained out of existence. So, if you feel so inclined, tell me down below what you think.
Remember that the world is made up of voices tiny and large. The tiny ones are not always tiny because they are not insightful, just as the large ones are not always so because they are the smartest. I think we need to find ways to help foster the tiny ones a little bit more, and I think A.I. can help us achieve that end.
LP