Algorithms and mental health
Mental health has been receiving a lot of media attention recently, and now it seems machines can be programmed to recognise depression in Instagram users. Past research has found that people suffering from depression prefer – and associate their moods with – darker colours; the work of two American scientists has now built upon this by revealing a strong link between mental health difficulties and the posting of images replete with grey, blue and black hues.
The study
Some 170 workers from Amazon’s Mechanical Turk service – 70 of whom were later found to be clinically depressed – granted Andrew Reece (Harvard) and Chris Danforth (Burlington) access to their Instagram accounts. The photos from these were then analysed for hue, colour saturation and contrast, measures that show how vivid a picture is. The number of faces in each image was also evaluated as a gauge of social activity, as well as the number of likes and comments on the pictures.
A machine learning algorithm was then employed to spot correlations between the image properties and depression. It found that the ‘Turkers’ with depression tended to post darker images, which received fewer likes, than those without the condition. It also recognised that the ‘Inkwell’ filter, which converts colour photos to black and white, was disproportionately favoured by those with mental health difficulties, while ‘Valencia’ – which lightens photographs – was favoured by mentally healthier individuals.
The machine was then put to work on the images posted by a further 100 individuals, 70 percent of whom it correctly identified as having depression.
A step forward
Technology such as this could prove instrumental in the identification and diagnosis of depression, which could in turn lead to individuals getting help in plenty of time. Since suicide is the biggest cause of death in the UK for men under 45, developments of the algorithm used in the study have the potential to save lives. Indeed, since the study, Facebook has begun testing an algorithm that identifies linguistic patterns suggesting a suicide risk.
The media exposure of the Reece/Danforth image study and Facebook’s recent algorithm tests will doubtless lead to greater development in the wider industry of software that can recognise people with mental health problems or who are at risk of self harm. Of course, such innovations would not be without their tax rewards.
Extensive research and development into this area is going on in the US, but UK businesses wishing to develop their own similar algorithms could benefit from a corporation tax reduction as a result. With the UK government keen to keep Britain at the forefront of innovation, both small and large businesses could reap rewards from looking into this technology further.
R&D tax credit for software companies
If your company is engaged in R&D and would like an estimate of how much tax relief you could be due, you can utilise our R&D tax claim calculator. Alternatively, contact us today on 0161 298 1010.
