Holy $&#!

This post is going to discuss how we at Myxer approach the issue of potentially offensive material appearing on our MyxerTones website, or otherwise being made available through our Myxer platform. Where do we draw the line with respect to what can be posted to the site? Given that we are essentially just a service provider for our user community, should we instead give tools to the community that allows every visitor to draw their own line? Should we feel obligated to help balance, e.g., the right of every person to self-expression with the wish of every person to avoid hateful, disturbing, or vulgar content?

These questions are very old, and there is no solution. Hopefully, this post will at least help people to understand how Myxer currently feels about some of these topics, and will provide enough background to demonstrate that we take these issues very seriously – probably far more seriously than a casual visitor to our website thinks at first blush.

Discussing potentially offensive topics without offending

Speaking of blushing, one of the problems I face in writing this post is figuring out how to talk with any specificity about profanity, nudity, drugs, guns, or any other fun stuff (ha, ha) without actually offending anyone.

Oops, too late.
Inside our company, I think we’ve gotten to the point of being able to communicate relatively efficiently about these topics. For example, our meetings on the subject no longer consist of each of us taking turns blushing, laughing nervously, or avoiding eye contact while gently lofting various uncomfortable words in an apologetic and barely audible manner. But it has taken time to get here. And were it not for the fact that all of us here are good friends with a high degree of mutual respect (and generally a long history together), this would’ve been almost impossible. If we weren’t so close we wouldn’t know, for example, that he’s not a pedophile and she’s not a racist and this guy doesn’t kill baby seals for fun on holiday.

Not much fun for little Harpo

Let’s get the obvious out of the way from the start: Everybody is different. Even the same person is different from one situation to another. So every visitor to our website is going to have a different idea of what’s offensive (and what’s fun), leading us to admit defeat from the get-go: it is absolutely impossible to insure that people are not exposed to something they find offensive unless they are exposed to nothing at all. Dory said something along these lines when she commented on a promise Marlin made to his son Nemo. Marlin promised that he would never let anything happen to Nemo, to which Dory said, “Well you can’t never let anything happen to him. Then nothing would ever happen to him. Not much fun for little Harpo.”

If we want to be fun (and we do), and we want to allow independent voices to express themselves in ways that differ from the prevailing “mainstream” (and we do), then we will sometimes offend some people. I’m sorry. Sortof.

Communities form, and form from, shared values

We all have values that are uniquely our own. They are, of course, shaped and shared in many very complicated ways by our family, friends, neighborhood, teachers, personal experiences, and probably a few hundred million years of evolution (if you’re into that kind of thing), among countless other factors. Part of what makes a community a community is, I believe, some set of shared values. So I think one might imagine it’s not out of the question that a litmus test could be developed that would determine, for some particular exhibit, for one particular community, whether it should be considered inappropriate or not.

It was this idea that led to the original method used on MyxerTones to identify what material should be considered inappropriate. What we did was add a “report as inappropriate” link next to every item, putting the job of identifying what shouldn’t be included in our catalog into the hands of the user community. (The original idea was that we would automate the process of removing inappropriate content from the site after it reached some threshold of “votes” by users, though we’re still doing this process manually).

What I really like about this approach is that, in principal, it allows us (Myxer) to remove ourselves from the position of being some moral authority. The community would use its shared values to determine what was allowed and what wasn’t, and we wouldn’t have to spend our time making lists of what were vulgar words and which configurations of human flesh would trigger an “inappropriate” label.

And if the Myxer user community had stayed a small one, and had intersected with a relatively small set of other communities, this might have been sufficient. The set of shared values might’ve stayed large enough to allow the “flag as inappropriate” scheme to be a fairly efficient filtering mechanism. But the Myxer community has grown to be larger than the population of many countries, and it attracts hundreds of thousands of people from very different backgrounds and external communities that – at least when it comes to determining what constitutes vulgarity or pornography or violence (or indeed whether these things are even offensive at all) – have a relatively small set of shared values.

This problem is just one inevitable result of the fact that the world is made up of an almost infinite number of overlapping, constantly changing communities of varying importance and influence. At one extreme exists “the world community”, a community that shares a relatively small set of common values. At the other, every individual can be thought of as their own personal community, with a rather more complete set of values. (I resist the notion that each individual has an entirely complete set of personal values, because it’s been demonstrated to me that even a single person often has multiple ideas that are put to use in different situations. Las Vegas has built a multi-million dollar advertising campaign around this idea (“what happens in Las Vegas…”).)

Divide or Dilute?

Faced with a large user community that has divergent tastes and values, the “flag as inappropriate” approach becomes problematic. It’s rather obvious in hindsight, but as we’ve continued to add people to our Myxer community, the percentage of contributed content that is considered “appropriate” by everyone has become smaller and smaller. This has forced us to consider whether we should maybe:

1. Lower the threshold of what is considered inappropriate to accommodate the values of the most (hard to find the right word here) conservative users in the community-at-large.
2. Define more than one threshold for offensiveness, along the lines of the MPAA ratings (PG, PG13, R, etc.)
3. Segment the population into smaller sub-communities that are more likely to share the same values, and allow each sub-community to set its own thresholds on inappropriateness;

The first option is essentially continuing on with the status quo. As our community grows to include people like advertisers who complain about images of butts-in-thongs, etc., we can crank down on our content to make it more and more sterile. This has the obvious effect of purging from the system a lot of things that are actually interesting to a lot of members of the community.

When your web community has a large number of 18-24 year old males, the probability that some of the images and videos they want to put on their phone and share with others could have subject matter that falls into the “potentially offensive” category for people who are not 18-24 year old males is pretty darn near 100%. At the risk of further offending a certain slice of my current audience, I submit that this last conjecture is virtually inescapable given the aforementioned millions of years of evolutionary biology.

The second option is actually already in use by Myxer in some fashion. We currently allow users to turn “content filtering” either “on” or “off”. When “on” (which is the default), content that has been flagged as inappropriate will not be shown. When the user chooses to turn it “off”, even inappropriate content will be shown to them. The main problem with the current implementation is that it’s only a binary thing, and it doesn’t account for people that want to, say, allow “profanity”, but still block…Julio Eglasias or something.

We’re actually in the process of trying to understand whether we can adopt some of the (admittedly very flawed) guidelines maintained by the MPAA (gag) or TV ratings people for Myxer. I don’t have terribly high hopes, but it might turn out that implementing something like this will at least keep people who don’t consider the issues as carefully as we do happy enough to give us their money for advertising.

Allowing sub-communities

The third option is philosophically really interesting to me, but it has implementation details that are difficult to wrap one’s head around. How does one choose the sub-communities? Is it possible to allow the sub-communities to self-evolve such that one need not define them ahead of time in a top-down manner? And then just the logistics of running an “appropriateness” test when displaying content from the Myxer catalog becomes computationally expensive and prone to error.

There are some specific segmentation use cases, though, that seem like they would pragmatically really useful. One thing we know from operating our website is that there is a rather large community of people who are really into hip-hop as a genre and as a lifestyle. The hip-hop community often makes use of language and imagery that, to people outside the community, is considered profane or hateful. The most often-cited example is probably “the n-word”, which we see used colloquially on Myxer in contexts where no offense is apparently intended.

Calling attention to the fact that I practice self-censorship when referring to “the n-word” above wasn’t my original intent, but it occurs to me that it illustrates the point fairly well. In my personal value system, that word is tainted to the point where I feel it is inappropriate for me to even spell it out – in my community, it’s just not acceptable. And yet I think it’s a word that should not be redacted in communities where it has acceptable and even very important connotations.

So I was trying to show an example of how segmentation of the user population into sub-communities could be useful. I’ll try to finish it off: if we had a hip-hop sub-community, that community could be allowed to self-police with the same “report as inappropriate” functionality we originally implemented site-wide. The ever-changing values of this hip-hop community could be tapped into to evolve what Myxer considered ‘inappropriate’ for people within that community. So if you were in that hip-hop community, it might be that there was a completely different set of words that were considered (by our system) to be profanity, which would prevent the values of other communities from restricting the speech of the hip-hop community.

Unfulfilling Conclusion

That’s really all the typing I can justify doing on the background of this topic right now. Needless to say, we at Myxer are extremely interested in doing The Right Thing (or at least The Rightest Thing Possible) to preserve the empowerment aspects of our technology while maintaining a comfortable and friendly atmosphere for people who visit our site. We want neither to censor nor to offend, but we understand these are impossible goals toward which we can only hope to ever advance.

We do have some specific refinements to our internal policies that we are putting into place that are too specific to warrant getting into, but this is just a small step and we will continue to adjust course time and again as we continue to grow and learn. I hope this discussion has been useful for at least one of the two people who made it all the way to end, and I would welcome any comments or suggestions you have on how Myxer can further improve.

Peace,

Myk