You know what IBM Watson is. But if the title of this article did not make sense to you, consider yourself lucky. You are lucky because you have not been exposed to or worse put through a highly questionable and non-scientific but popular method of personality test called Myers-Briggs.
Your luck however just ran out as you are reading this. M-B test classifies people into 16 groups labeled by a set of four letters with 4 four possible options for each letter. They do that based on your answers to a personality test. Many researchers have written in-depth about the flaws in this test. You can find those here, here and here.
Simplest objection to Myers-Briggs is this is not scientific. That is to say it did not start with a hypothesis, did not do controlled experiments, did not account for limitations of methods and its methods, and its results are not reproducible. That does not stop businesses and managers from using it, making their people take the test and even asking them to post the four letters on cubes for everyone to see.
Many who took M-B tests and read its findings will strongly stand behind its findings, stating that the results are an accurate description of them. The text in M-B findings are meant to evoke those feelings, written in generic enough and highly positive language that most will see as description of them.
There is indeed an insatiable interest among us to measure personalities, to get a view into our minds, to see who we are, to know more about our employees, customers, followers etc. Tapping into this interest IBM Watson introduced a new Personality Insights tool.
The IBM Watson Personality Insights service uses linguistic analytics to extract a spectrum of cognitive and social characteristics from the text data that a person generates through blogs, tweets, forum posts, and more.
Unlike the lengthy list of either-or question based tests you take for M-B and other tests, with Watson’s test you cut and paste your blog postings, public writings, tweets, emails and anything you willingly shared with others. Based on text, sentiment, pronouns and linguistic analysis, Watson tells you who you are.
At first take, I expressed my skepticism of this method. I am changing my stand here, not because I find the insights to be an accurate description of me but because I spent time reading their detailed description of methodology, data and limitations. In other words the science behind this.
I do not know whether the findings are actionable but here are the key aspects of the study that tells us this is scientific and on the right path
- Started with a hypothesis: Language reflects personality, thinking style, social connections, and emotional states. The frequency with which we use certain categories of words can provide clues to personality, thinking style, social connections, and emotional stress.
- Analyzed prior works: There is a lengthy survey of prior scientific works and understanding of what they did well and the limitations.
- Adding a verification step: The common problem with other personality tests that claim they are scientific is they use the very data used to build the theory as evidence of its proof. On the other hand IBM conducted a set of studies to understand whether personality characteristics inferred from social media data can predict people’s behavior and preferences.
- Transparency in models and inference: There is no Harry Potter sorting hat here nor any axiomatic M-B like statements. IBM shares the complete model and an explanation of how inferences are arrived. All the assumptions are shared for any researcher to test.
- Understanding of limitations: The work lists no less than seven significant limitations with the method and its results. Compare this to how M-B practitioners and fans speak about the test and its findings with willful suspension of skepticism.
Does this test work or is it relevant to you? I do not know, but at least you know now the level of rigor needed before you pay up big fees for personality tests. IBM Watson’s Personality Insights tool gives you the ammunition to dismiss unsupported and unscientific personality tests.