Of late, privacy has been in the news. For example, Google Glass is being questioned over how it would affect privacy. A cafe in Seattle has already banned it.
Actually, privacy has always been an issue since the rise of social media . We’ve all read the stories, but that’s not where I’m going with this post.
We talk about assessing social media performance. But it’s not like we’re assessing a motor engine. We’re assessing the activity, the interactions and even the sentiments of what sentient people are saying.
We’re assessing people and they probably don’t realize it. Our dashboards are saying so-and-so are “influencers” because the system looked at their public blog, Flickr and Twitter accounts and see that there’s a connection and they talk a lot about our brand. Fine if it’s positive, troublesome if not.
I might be over-reacting. Of course when people praise or complain about a brand, there’s a good chance they do want to be noticed. But for them it’s a sort of social, transient interaction on the surface. I bet most of these people do not realize many companies and their social media vendors are also putting down their postings and identities in reports to management. For good, or for whatever.
Did I say it’s ok to assess me?
As social media enthusiasts and workers in this field, we sometimes forget we are also being archived and examined. This post will soon be visited by search engine bots and archived at some point, for example. Personally I don’t mind that, I even intend it. I’m consciously aware it will happen. Whereas many people out there don’t think about this aspect of being on social media – that you’re not just being read on the page, but also turned into insights, analysed for sentiment, counted for virality and even put into monthly reports.
The standard argument is that – you posted it in public, so it’s free for companies to data-mine. Is it? Maybe the terms are in the TOS I never read. :)
Is 70% of the truth, truth?
One question worth pondering about is the issue of sentiment analysis accuracy. To date, some 4-5 years++ after sentiment analysis became active in mainstream businesses, the accuracy remains less than 100%. Much less in fact. Some recent writers suggest even 70% accuracy is considered an optimistic ratio, and that is the realistic ceiling.
I personally think 70% is not only not accurate, but it’s also just a number. A number trying to measure the accuracy of something trying to objectively measure something subjective. In other words, I am highly tempted to say this number is just as airy.
OK, but that’s actually not the point here. The point is this: if say sentiment analysis and other social media measurement tools are not entirely accurate, just how much of the assessment of what people say or do online is “accurate”?
What if a sarcastically positive tweet I made is interpreted as positive in sentiment when I meant it to be negative, and this is reported in some brand’s report to management as “positive”? And product/service decisions are made based on this reading? The end result is that the company might believe it is doing a good job, whereas customers do not think so. Is this not unlike how some middle managers filter their reports to upper management to make it look like they’re doing a good job?
Sentiment Analyzing Self
Let’s bring this back to us: so we who have to prepare social media assessment reports to our managers need to take care. You are not just handling computerized insights – you are handling real people, and their sentiments. You are an agent of sentiment and veracity – what you report can be construed as putting words in people’s mouth and that’s a huge responsibility. So, when you put on your assessor glasses, don’t forget to assess yourself.