Personality Profiling Systems give a very unique view of an individual’s personality. However, let us be careful to properly establish what it is that we are looking at? There are many differing perceptions of personality and not all the many definitions of personality can be assessed. By using psychometric personality assessment we are describing how an individual might react to a given situation or how they behave in a given setting. How they behave or respond in a public or group setting (Persona) or how they are in private. What their preferences and likes are or what kind of environment suites them best. Often this technique is used to assess suitability for a job role in recruitment, but it is also a valuable tool in the development of individuals and teams.
In this article I set out to explain the thinking and process of how a personality questionnaire actually works. For the benefit of the reader most of the complex technical detail has been omitted.
Let me start at the beginning of the process by discussing the questionnaire, it’s purpose and development. Why do we use a questionnaire and how are they developed. The questionnaire doesn’t have to be a written document, it can be a conversation or interview where the interviewer asks a set of structured questions of the interviewee. This is not the most cost effective method of getting the information required, nor does it yield consistent results, however, in certain circumstances it is still used. In preference we tend to use a questionnaire that is either paper based or online. To get the information about someone’s personality we ask a set of structured questions, just as one would in a typical social gathering. For example, the host when offering the canapé’s to someone may ask the question “do you prefer cheese or salmon?” Our questionnaire attempts to mimic this process by asking questions that have a choice of answers and our questionnaire consists of 225 such questions. The vast majority of the questions are asking what the respondents preference would be, or how they would respond or react to a given set of circumstances or scenario. In our personality profiling system we have seventeen traits or scales and all questions are tied to one of these scales. The questions for each of the scales are randomly distributed throughout the questionnaire. One could argue that it is really seventeen separate questionnaires all melded into one. As all of the scales work on the same principle, we will only look in detail at just one of the scales, the one that measures ‘warmth’. This trait is assessed or measured by countless personality assessments the world over and whilst it would be novel to invent a completely different set of criteria, it wouldn’t be very helpful as we are attempting to assess the human psyche and as we are all essentially the same, it pays dividends to stick the same criteria as other test publishers.
So Warmth, how do we assess warmth? The first process is to establish a scale and we can do this by saying that at one end of the scale, the individual isn’t warm, and at the other they are warm. Another way is to establish a scale is where one end is cold and the other end is warm with a mid-point of not warm or cold. This second method allows us to measure two things ‘warmth’ and ‘coldness’
So let’s describe what this scale is assessing, what do we mean when we describe someone as ‘warm’ or has ‘warmth’ and what do we mean by ‘cold’. One way to discover this would be to ask a large group of people to describe ‘warmth’ or ‘coldness’ by thinking about people they know. Typically the answers would include ‘likes people’, ‘joins in with others’, ‘friendly’, ‘genuine’, ‘a nice person’, etc. Coldness would elicit ‘aloof’, ‘remote’, ‘distant’, ‘difficult to get to know’, etc.
So having established a scale we now need to design some questions that when asked they reveal whether the individual is warm, cold or neither. In designing the questions we need to think about how the person asked is likely to respond, we can ask someone if they are abrupt or aloof, but just as in a social meeting, the conversation would very soon end if the questions were that blunt and insensitive. Also, just as in a real conversation the respondent would not answer the question or answer in a more favorable light and consequently the response to the questionnaire would not be very accurate. So we need to ask questions in a much more subtle way. In the social conversation we would ask ‘how do you like meeting people’ to which the respondent could reply ‘Oh no I dread that and keep out of the way’ or ‘Yes, love to meet new people I find them fascinating’ and so by creating little scenarios or asking the respondent whether they agree or disagree with a particular view we can gradually build a picture of where a person sits on the scale. We do need to ask quite a number of questions that build up a picture of the individual that is answering the questions, we also need build in some redundancy for the questions that simply do not ‘fit’ the individual.
We often refer to the psychometric properties of a question and this simply means that the question will reveal, when answered, an indication of the mental state of the individual, for example the question might be “do you feel sad when others are dishonest” this would reveal the metal reaction that the individual would feel when faced with someone being dishonest, as we are trying to assess someone’s personality then all the questions we ask require a psychometric property.
There are a number of ways of recording the results, one common way is to ask the respondent to rank on a scale of one to ten whether they agree or disagree with a given statement within the question or another option is to give two opposite scenarios and ask the respondent to choose which one (or neither) this method will give a yes, no and neither option, which is useful for polar scales. So having designed a set of question and a scale to measure the responses against we now need to adjust the scale to make it accurate. Let me explain what is required, when we describe someone as being warm or cold we are assessing them against our perception of what we judge to be warm or cold and we do this by comparing them to others. For example we may hear a comment from another student along the lines of “you will like the new teacher as she is very warm and engaging” on the other hand this might also be heard “the new teacher is a bit aloof, but she knows her stuff” if the teacher was neither warm or cold then it’s unlikely the student would need to make a comment at all. We only tend to comment when the individual is outside of the normal expected range as this is something we would notice. For example if it suddenly starts snowing here in Brighton, it would draw comments from people of “Oh gosh, it’s snowing” however, the same situation in northern Alaska is unlikely to raise an eyebrow. So how do we decide what is ‘normal’. Normal is and average of what we expect for a given measurement. For example a normal daytime temperature for August here in the UK could be 21 degrees Celsius, so how do we decide that? One way would be to take the temperature every day in August for many years and then average all the collected readings, this would give us the normal expected temperature for August. When the temperature is much higher then we can say its above normal or lower, then its lower than normal.
Let’s now apply this principle to our questionnaire. We ask lots of people to complete our questionnaire and then we average the results and this gives us the midpoint of our scale. We can then decide those who score above the midpoint are warmer and those who score below the midpoint as cooler. The midpoint being what we would expect as normal or the ‘norm’ as we refer to it. Whilst the actual mathematical process for establishing what we describe as a ‘norm’ is far more complex than described here, this is, in principle the process used.
As already mentioned, the establishing of a ‘norm’ is a complex mathematical process but if the questionnaire is to yield accurate results then it is an essential one. There are some fundamental principles of statistics that we need to employ here and ‘normal distribution’ is one of them, it is a natural phenomenon that has been challenged over and over again through the centuries and its simply that 68.2% of the population will fall within the normal range of -1σ (Sigma) to +1σ (Sigma) or one standard deviation from the norm. So anyone scoring within this range would not be described as either warm or cool.
How do we achieve this ‘norm’? The norm is obtained by applying a numerical value to the question answers, typically a ‘warm’ answer would score 2 and a cool answer would score 0 with a neither answer scoring 1. The next step is to add all the scores from the completed questionnaire together and then divide that result by the total number of questionnaires completed, this gives us the average score or ‘mean’. However this mean might not be true, let’s explain why, let’s say a hundred school children undertake an arithmetic test and the vast majority scored 45 out of a hundred but nine children scored 89 out of a hundred. This would make the average score nearly 49. So by this simple calculation ninety one children scored below average! Hence my comment that the mean might not be true. In this situation we have nine children whose scores have ‘deviated’ from the average by quite a long way. In this situation the deviation is large and easy to spot so the teacher could treat the nine children’s results differently in an attempt not to skew the overall result. However, in the real world the deviations of individual results from the mean are much smaller and we are dealing with a much larger number of questionnaires, so spotting them by eye would be an impossible task. To overcome this problem we need a mathematical solution to establish what the standard deviation of our data set is.
We use a mathematical calculation that works out the standard deviation:-
- Find the mean.
- For each data point, find the square of its distance to the mean.
- Sum the values from Step 2.
- Divide by the number of data points.
- Take the square root.
By using this process it will give us the standard deviation for our dataset and we will see that 68.2% of the respondents will fall into the -1σ to +1σ range. We apply Bessel's correction to calculate the standard deviation for our data set as it gives a more accurate result.
A further calculation with an individual’s score will show whether it is above or below the standard range and if, for example the score is +1.5σ we can describe this person as being a ‘warm’ individual who will like being in the company of others, sociable and gregarious. This is achieved by converting the individual’s raw score (the sum of the answers from the questionnaire) to a ‘Z’ score by subtracting the mean score from the raw score to yield a difference score and then divide this result by the standard deviation. This score will be in the range of -3σ to + 3σ. A much more convenient score to use is a standard ten score (STEN) and this is achieved by multiplying the ‘Z’ score by 2 and adding 5.5. By hand these calculations are very long and tedious, especially as our main questionnaire has seventeen traits and each one would need to be calculated separately, so thankfully we have a computer program that completes this task in seconds.
There you have it, when we measure ‘warmth’ we are making a subjective assessment of how warm an individual is compared to a group of other people. Of course, the accuracy of this assessment will depend on the size of the pool of completed questionnaires that have been used for the ‘norm’. The bigger the pool the better the result. Comparing the individual against norms derived from different professional groups is worth considering as it would be a fair comment that the vast majority of sales personnel are likely to be warm and outgoing so having a norm of just sales people would indicate which ones would be warmer than the average and the convex is also true, someone who is cold compared to this group is unlikely to succeed in this occupation. Medical professionals tend to have a very high attention to detail (thankfully!) and this is a trait that we can establish using a ‘norm’ and this then creates a benchmark of what an individual is likely to behave like.
By getting many successful professionals within an industry or profession i.e. Accountants, Doctors, Lawyers, etc. to complete the questionnaire we can establish a norm and this would give us an indication of potential success of an individual who wants to work in that profession.
So the next time you are faced with the prospect of completing a personality questionnaire you might not meet it with such dread.
Malcolm Yates MSc, Chartered FCIPD, FInstLM, MPsych