Personalization? Get Out of My Head

There is lots of web chatter around on sites such as LinkedIn and Twitter on the promise of personalisation. Artificial Intelligence (artificially intelligent? ;-)) applications allegedly allow us to determine several constructs which may support the differentiation of products, services and messages according to individual differences. These include factors such as mood, sentiment and even social linking strategies.

But new methods of data collection don’t make a new discipline, and tagging a weak model with an AI headline doesn’t make up for the weaknesses.

After all, interest in personality and individual differences isn’t new.

Since Allport and Vernon‘s assessment of the field in 1930, structured research into personality and individual differences research has expanded enormously. This is particularly so in terms of theoretical development and practical relevance, where the research is used for both assessment and application.

This has led to a number of challenges. From the viewpoint of assessment, we can see concerns. The idea that individuals may differ naturally and meaningfully — particularly in abilities — challenges key ideologies and has shaped some of the most tragic episodes in human history.

Indeed, research into individual differences research often claims to establish that there are robust, inevitable, and salient differences between and within people that are partly heritable and have long — term consequences. This leads to all sort of issues around privacy and discrimination.

Traditionally research into individual differences aimed to help us understand how and why individuals vary in their affect (moods and emotion), behaviour, cognition, and motivation. To this end, researchers in the field have attempted to accurately describe, explain, and measure dimensions of individual differences, to evaluate the long — term consequences of such differences, and to discover their aetiologies, including their biological, environmental, and genetic bases. They have employed a wide range of research tools and theoretical approaches, spanning psychometrics, brain imaging, as well behavioural and molecular genetics to create varying models of individual difference.

This has led to a large -some might say overwhelming — body evidence demonstrates that individuals differ along continua of affect, behaviour, cognition, and motivation, many of which are understood and operationalized in terms of trait dimensions based on structured

And there is yet still no consensus.

From the viewpoint of application, there are even more problems. There is, within the domain of personality research only a loose consensus on models and terms within and across different research communities.

Current proposed uses of personalisation data are to adapt and target content in the form of advertisements, opinion pieces, and the widest possible range of other message forms. These customisation efforts may support changes in affective, cognitive and motivational content elements which will conform to the individual differences.

But doesn’t personalisation require a global model of these individual differences? After all, if we all employ a different model of individual differences how can we validate it? Which aspects of our ‘person’ will correspond to which behaviours? Which aspects of our behaviours correspond to which aspect of the ‘person’?

And what happens if the factors which create the differences are fluid?

Everyone who interacts regularly with other human beings, recognises that issues about their assessment, consequences, and aetiology is sensitive and controversial. If the AI being used to modify content, and the data models which it uses are not completely accurate, users will be particularly unforgiving, and in the event that such data is hacked, the possible class action might prove very expensive indeed in terms of lost revenues and lost user base.