How to know your target (audience) better than they know themselves

Monday August 1st, 2016

An incredible amount of information about a person hides within the language they use, and surprisingly the way in which language is constructed can be significantly more insightful than the words themselves. Through over 20 years of research in social psychology, Jamie Pennebaker (University of Texas Professor and co-founder of new Teradata partner Receptiviti) discovered that it was the ‘function words’ that we normally discard when performing text analytics that are key to understanding the person behind the language. For example, how often a person says ‘I’ provides insight into how honest they are being, how depressed they are, and also whether they are communicating with a superior or inferior audience (see Jamie’s awesome TEDx talk).

Receptiviti has taken Jamie’s research and productised it into API form, providing the capability to understand peoples’ psychology, personality, thinking style, emotion, tone and relationships by analysing the language they use (whether it be text they have written or transcripts of spoken conversation). Given as few as 300 words, the Receptiviti API will return a personality profile comprised of 130 numeric scores for measures in six broad categories: ‘Cognitive/Thinking Style’, ‘Big 5 Insights’, ‘Social Style’, ‘Emotional Style’, ‘Working Style’, and ‘Interests and Orientations’. The measures themselves include ‘Openness’, ‘Family Orientation’, ‘Depression’, and ‘Type-A personality’, as well as ‘Sexual Focus’ and ‘Religious Orientation’. The resultant profiles are reminiscent of the personality questionnaires that we all know and love, but with a significant difference – the subject is not self-reporting on their own impression of their personality, making it incredibly hard for them to game the system.

Sean Farrell - Target Audience 1

When I first stumbled across Receptiviti and their wizard-like ability to see the personality behind the language, a number of applications immediately jumped to mind. Imagine being able to understand your customer like never before, have the ability to know exactly how best to engage with them and be able to identify which of your staff they would best relate to. Or what if you could measure the satisfaction/dissatisfaction of your employees in real-time and be able to predict who is likely to leave, giving you the ability to intervene before you lose valued talent? Understanding people has enormous applications across a broad set of use cases, from Marketing to HR to national security. However, building a personality profile is just the first step – you then need to know what to do with it. And that’s where advanced analytics comes into the picture.

At the time I was first introduced to Receptiviti and Jamie’s research, I was playing around with text analytics and social media data. I was experimenting with classifying samples of tweets using machine-learning techniques and examples of tweets that were of interest versus those that were not. While I was having quite a bit of success with my models, I couldn’t help but think I was only scratching the surface – I could easily find content of interest, but what I was really interested in was finding people of interest. Could I use Receptiviti’s personality profiling capability to achieve this aim?

Sean Farrell - Target Audience 2It turns out the answer to this question is a resounding ‘yes’. The numerical measures in the personality profiles are perfectly suited to feeding into machine learning algorithms, and provide a powerful data source for seeing way beyond the text and language to uncover the people beyond. I played around with profiling a number of well-known individuals, and was very successful at building models to classify people as belonging to particular similar social groups. I was particularly good at finding Richard Branson using his social media and blog posts in a mass of data from other users (much like a high-tech version of “Where’s Wally?”).

Applied in the national security domain, this has shown exceptional promise in identifying members and supporters of violent religious groups. If susceptible, disenfranchised individuals can be identified as being on the path to radicalisation, this would provide a massive opportunity for community leaders and law enforcement officers to intervene early enough to gently steer risky individuals in a different direction (a very hot topic for many government departments focused on countering violent extremism). Other applications in this area include flagging potential perpetrators of mass shootings in the US, identifying members of criminal organisations (such as people smugglers or drug traffickers), or tracking down members of pedophile rings.

Such technology does of course raise many issues in the debate of privacy versus security, and it should be strongly stressed that there is no silver bullet solution that will provide definitive classifications for ‘good’ versus ‘bad’. Instead, the combination of machine learning (vital for scaling to meet the growing task created by the increasing data deluge) and personality profiling can provide a triage tool for security officers to better focus their resources and attention on the most likely threats. It’s then up to the expert analysts to apply their own experience and skills to determine whether the person truly is of high risk to the community. In light of the spate of recent terrorist attacks (in Baghdad, Istanbul, Orlando, and Nice among many others) the pressing need to identify potential terrorism actors in advance is clearer than ever.

Of course, the applications go way beyond law enforcement and national security. The same techniques could be used to build models to seek out the consumers most likely to be interested in your product or service at a level of granularity that could previously only be dreamed of. Or to sort through hundreds of job applicants to tease out the candidate who is perfectly suited to the role a HR manager or recruiter is trying to fill. Or to churn through millions of online dating profiles to find that one true love (again, see Jamie’s TEDx talk to hear how he was able to predict speed-dating matches with a much higher accuracy than the participants themselves).

The combination of personality profiling and advanced analytics techniques really does provide boundless opportunities for understanding your target audience in the era of truly Big Data. All that’s needed is the imagination and creativity to recognise the potential.

The following two tabs change content below.
avatar

Sean Farrell

Data Scientist at Teradata
Sean Farrell is a Data Scientist with the Teradata Australia and New Zealand Advanced Analytics group. Based in Canberra, he is responsible for providing analytics expertise and support for Teradata engagement with the Australian federal government. Sean has 11 years of experience working in analytics, primarily gained through his time working as a scientific researcher in the field of astrophysics. Prior to joining Teradata, Sean worked as a data scientist for the Australian Department of Defence. Before moving to data science he worked as an astrophysicist in France, the UK and Australia. He has worked for the European Space Agency on the XMM-Newton space telescope mission, discovered a new rare class of black hole, and observed with the Hubble space telescope. He is internationally recognised as a world expert in data mining astronomical datasets utilising advanced analytics techniques including machine learning. Sean has a Bachelor degree in Physics and a Bachelor degree in Mechanical Engineering with Honours from the University of Newcastle, and a PhD in High Energy Astrophysics from the University of New South Wales (through the Australian Defence Force Academy).
Category: Sean Farrell Tags: , , , , ,
avatar

About Sean Farrell

Sean Farrell is a Data Scientist with the Teradata Australia and New Zealand Advanced Analytics group. Based in Canberra, he is responsible for providing analytics expertise and support for Teradata engagement with the Australian federal government. Sean has 11 years of experience working in analytics, primarily gained through his time working as a scientific researcher in the field of astrophysics. Prior to joining Teradata, Sean worked as a data scientist for the Australian Department of Defence. Before moving to data science he worked as an astrophysicist in France, the UK and Australia. He has worked for the European Space Agency on the XMM-Newton space telescope mission, discovered a new rare class of black hole, and observed with the Hubble space telescope. He is internationally recognised as a world expert in data mining astronomical datasets utilising advanced analytics techniques including machine learning. Sean has a Bachelor degree in Physics and a Bachelor degree in Mechanical Engineering with Honours from the University of Newcastle, and a PhD in High Energy Astrophysics from the University of New South Wales (through the Australian Defence Force Academy).

Leave a Reply

Your email address will not be published. Required fields are marked *


*