Split-Personality Insights

I started my day today by completing the IBM Personality Insights test which analyses your personality based on your last 200 tweets. Before I even completed this test I was skeptical, which is ironic for a reason you will find out shortly.

For context, I believe my online presence is a fair and largely honest representation of who I am personally. I have one Twitter account and one LinkedIn account. LinkedIn I keep strictly professional, but I fully embrace Twitter and those connected to me will know all about me, mainly my profession and related thoughts, my family life and my hobbies and my love of Boxing and Swansea City FC. But I am also aware that in this modern world its sensible to ensure a positive reflection of oneself is portrayed online as my career is important to me, so I hold off on the bad language, I never ‘drunk-tweet’ and I do choose to keep things to myself that should be private.

So here is what the report said about me;-


Here’s what I agree with;-

  • I’m skeptical (I prefer discerning, but i’ll take skeptical)
  • I am reasonably authority-challenging
  • I am philosophical, I am open to and intrigued by new ideas and I do love to explore them.
  • I care more about making my own path than following what others have done.

Here’s what I disagree with;-

  • I am not inconsiderate (this is contrary to every other personality test i’ve ever completed)
  • I am not tranquil (but boy, I wish I was!)
  • I am not generally serious and I do my fair share of joking around.
  • I am not really bothered about prestige.
  • I am very concerned with helping others – this is so hugely important to me.
  • Perhaps most controversially, I absolutely love Country Music (my favourite band is The Band, I love Merle Haggard, The Carter Family, The Byrds and don’t get me started on the Flying Burrito Brothers!)

There are one or two other comments that I neither agree nor disagree with but the above is the general overview.

I must admit that I was taken aback by the thought I could be perceived as inconsiderate and unconcerned with helping others and I felt almost offended by the results. In fact, what I should have done, is read the small print…

I believe (although I can’t validate this as I can’t find the supporting material) that it reviewed my previous 200 tweets, so this probably covers a period of around 3 weeks, during this time I watched my beloved Swansea City lose, a boxing match that genuinely disappointed and angered me, and a government budget that didn’t sit comfortably with me, so this was probably a ‘heightened’ period for me which was reflected in my tweets.

So I have a range of questions from this so far.

  • Is your SoMe profile a fair reflection on which to take a personality test?
  • Is 200 tweets enough to gain a proportionate view of your personality?
  • How does the tool consider your own tweets versus the re-tweets and quotes you make of other tweets?
  • How does the test detect humour and sarcasm?

Fundamentally, this brought me on to this question.

Am I looking for ways to disprove the accuracy of the test because I don’t like what the test said?

And as a follow-up…

If the test has this perception of me based solely on my SoMe presence, do other users have this perception of me too, and does this mean my presence needs TLC?

This second question is more worrying largely because I don’t want anyone thinking I’m not helpful or considerate, and I certainly don’t want anyone thinking I dislike Country Music!

I believe human users can look beyond these personality insights and form insights based on sense, emotional intelligence and existing knowledge whereas a test such as this IBM Watson test, is essentially ‘Garbage In, Garbage Out’. Yet, this test has still left me uneasy, and somewhat paranoid.

What this test has taught me, is whilst I have a fairly newly adopted SoMe open-book policy, if users share this perception of me or run similar tests on me in the future, then its very possible that I need to be more careful, and need to be proactive on this front too. I’m not saying I need to change my personality, but need to understand how tests such as this work so that the results are a more accurate reflection of the real me.

I suggest you complete your own test, and let me have your opinion on your own test results. The link is here -> http://gwen-systemu.mybluemix.net

This post is a to be continued… this needs more thought.

3 thoughts on “Split-Personality Insights

  1. The IBM Watson personality insights stuff is a really interesting area, Mark, and you’ve hit on a number of areas that need to be looked at carefully before it’s genuinely applicable for HR purposes.

    I’ve seen a number of people sharing similar reflections on just how reliable and credible this kind of analysis actually is and the 200 tweet limit certainly plays a part, I’d suggest. There’s actually a better implementation to try out that goes further than just the 200 tweets. You can try it out here https://personality-insights-livedemo.mybluemix.net/

    There are some other things to consider.

    As far as I can tell, the IBM Watson service simply outputs numerical values across the various personality dimensions. Someone has then written a really basic series of canned statements and tied them to certain values that are provided. I think there’s a high probability that that’s one significant source of error in what you’re seeing.

    Then there’s the way that we use Twitter. Let’s face it, those of us who use Twitter for work are in a minority. Take a look at most surveys and analysis of what people use Twitter for and we’re outliers. That’s important because IBM will base their analysis on “normal” twitter usage. The number of assumptions that are inherent in the Watson analysis means there’s scope for error there too. IBM even states: “IBM believes that, ideally, the text that is used to infer an individual’s personality needs to be reflective”. But there are other stipulations and assumptions there too and this document makes for interesting reading https://www.ibm.com/watson/developercloud/doc/personality-insights/guidance.shtml.

    This stuff is clearly experimental, fine for testing out on marketing tactics, less ready for high-impact, personal use. It’s fascinating though, there’s huge potential there.


    1. Hi Owen, thanks for reading and your comments. Your explanations and observations have opened up the tool for me and i’m glad others have the same experiences and concerns with it. I’m going to try out the broader test that you linked to and see how that differs. I’ll let you know how I get on.

      Thanks again Owen.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s