There is a growing challenge to retain control of the identities we build up online. Bias is alive and kicking even without the presence of machine learning but once we start using those AI methods more frequently, it will only make the problem worse. The networks and platforms we entrust our data with could (and should) help us more in the way we share our personal information with the world.
Our behavior and that of others is affected significantly by how our identity is perceived in a given situation. For example, if we believe we are anonymous, we’re more likely to take risky actions as we expect the consequences to us would be low. On the other hand, if we think we have a reputation to uphold, we may pander to the audience or feel the pressure to live up to expectations. Other people’s behaviors in the same way will vary depending on who they think they are talking to. If you’re perceived to hold a high professional status you’re likely to experience a different response than a trainee, even if you say exactly the same thing. This type of biased behavior is just the way humans react.
In every online network we sit on an identity scale anywhere from anonymous to famous (or even infamous) plus all the levels in between that are built up based on what we give away about ourselves. It’s just as true in real life as it is online but being online gives us the added opportunity to easily adjust or disguise our identities in order to project the image we want others to see.
We have so many factors that make up our perceived identity e.g. username, real name, gender, age, profile picture, history or bio, status, number of followers, group affiliation, religion, what actions we took, and more that can be casually shared without paying much attention. You may be happily sharing your status as a parent on social media without too much thought, but a segment of people may be, unknowingly to you, making strong judgements just based on that one disclosure. The list of factors that make up identity is endless but somewhere in there is the one factor you think people are paying attention to — what you said or the content you posted. You hope that what you say or do is the thing that is most important but it’s probably a hundred other things that are determining the response (or lack of) that you get, most of which are out of your control. Even if you tried to be anonymous by giving away just a username, you’re not safe from bias that way either. Being unknown means you fall victim to the opposite problem i.e. no-one wants to hear what you say if you’re a nobody, they need to know more about you before they care. Giving away too much or too little is a tricky path to navigate.
Whatever you do online, there is judgement, bias, and actions taken in response to things you have no control over or didn’t want to be factored in. They are related to your perceived identity and it’s often far removed from your actions or communicating the wisdom you might be expressing at that time.
What role have our social networks and platforms played in helping us to navigate identity?
Social networks aimed at family and friends want you to share everything about your identity, professional networks revolve around status and history, while interest based networks require little more than a username to get involved. In each network, you decide whether you want to sit on the anonymous end of the scale or strive for the famous one. Except, you can’t be sure what your data is being used for and how it is messing with perceived identity in that network — the clarity on what personal data is being presented to others or within the system is rarely there. The promise of being online was that you could create a place that brought with it equality and opportunity to build a level playing field. Anyone could join in the conversation and bring their views to even the most disparate fields that they normally weren’t able to access. In some ways we are succeeding by allowing to networks so freely but the identity we portray online and the bias that comes with it seem to have made the problem worse for humans. Getting equal access is one thing but being treated equality once we’re in is another. This was a problem in all online platforms and existed before we started talking about AI bias. While technology companies are looking to install ethics councils to help solve the problem of bias with machine learning algorithms, we could take steps today to make it better.
Aside from the networks where your data is shared with other people, the platforms that serve us content or provide services are building up identity based on the factors you transmit knowingly or unknowingly. Instead of a scale that goes from anonymous to famous, the platform is placing you on a scale from unpredictable to predictable depending on how much they know about you. The same types of identity factors and consequently the same potential for bias is being created whatever the platform is for, as long as you are feeding in your data.
A further complication for those platforms or networks consuming data, is the purity of the identity. It turns out, a large portion of us will share our accounts for services like Netflix, Uber or Amazon. It could be within the family or close friends but the fact is many different people could be using the same account, each with their own preferences or usage habits. Algorithms are likely to struggle to find meaningful patterns when in reality, there are a number of people behind the scenes making decisions under the guise of one identity.
How could we give people more power to control their identity online?
If we were to start from scratch designing platforms again, this time with the aim to give back control of identity to the users and put a bigger focus on the content, we might do things differently.
For example, we could decide to leave out identifiers that aren’t needed. Why do we ask for profile picture or unnecessary personal information? We could offer a neutral, unique, visual representation that we automatically generate (not just a bland placeholder) rather than always put the onus on the user to share a photo and bio before anything else.
- We could stop the trend to pull in all kinds of personal data when it’s not of benefit to the user and only serves to conflate numbers. Do we really need to port in all the various profiles from other networks or pull in phone based contact lists — it’s not necessarily related to the type of content you want to share to this new network so why encourage it?
- We could make efforts to amplify the content more than the user. Identity details could be protected by presenting a narrative of the user that’s appropriate for the content e.g. in a professional network, I might want to share the highlights of my experience but not my photo or gender when I post a comment.
- We could give more control to the user to hide identifiers in some situations. How about an ‘identity slider’ that a user can turn up or down depending on the content being shared. As a member of any platform or network I may want to be flexible on the level of anonymity applied. For example, on a trending platform, I could be interested on seeing the trends displayed to the population from an anonymous standpoint and not just those ‘recommended for me’. In some cases you aren’t even aware the trends you are shown are only for you based on your identity and not the same trends everyone else sees. It can skew your perception on what is really trending in the world which is not a good thing.
Some of these problems could be solved in part if we created multiple identities on the same platform or network but why do we provide this as the only option? People change, their preferences and usage patterns change over time, and ultimately identities can and should have the ability to be modified to reflect the user and how they are perceived.
Identity needs to be treated with more sophistication than we have allowed it to be online. As we see more networks emerge that are based on trends, more user identifying information is being fed into machine learning models that are the building blocks to artificial intelligence. Our perception in the eyes of other people (or machines) will be even less in our control when that happens. Being in control of our identity is important and choosing what and when to share about ourselves should be something we have more control over in any online situation. The platforms we use should help us do that.