Googling Depression: the Risks of the Search Engine’s Attempt at Diagnosing Mental Illness

A couple of weeks ago, Google announced that they were partnering with the US National Alliance on Mental Illness to provide a new feature: whenever users would google ‘depression’ on a mobile device, the search engine will return right on top of the results page a link prompting users to take survey to see if they are depressed. This survey is based on the PHQ-9, a questionnaire which is widely used by GPs and physicians to check for depression and is based on the DSM-IV criteria for depression. The purpose of this feature is to raise awareness of mental illness and to encourage people suffering from depression to seek support. This is just like how Google took measures to prevent suicide by directing users with searches on how to kill themselves to appropriate helplines and resources. On the surface, everything seems well intentioned: Google is attempting to disseminate information about mental health, information that has some clinical basis, and they are similarly cautious to avoid overstepping their remit. The point of this survey is not to provide a diagnosis, but, according to Mary Giliberti, CEO of NAMI, a more modest attempt at helping patients ‘have a more informed conversation with [their] doctor’. But just like my previous remarks about machine learning and Instagram feeds being used to diagnose depression, I remain deeply sceptical of any such attempt at a tech firm like Google to offer such tools for diagnoses.

The immediate worry is that data provided to Google can be used for advertising. That is the most immediate criticism that the announcement has drawn from a number of commentators: like from John Grohol on the PsychCentral blog or Arwa Mahdawi on the Guardian’s Opinion portal. Google claims that it will respect the privacy of user’s responses and not store this data, nor will it use any of it for advertising, but Google’s reputation on personal information is far from trustworthy and critics, including myself, remain unconvinced. Even if Google does not use any of the responses to the questionnaire, they would have already tracked a user’s search for ‘depression’ and profile users accordingly. But setting aside this issue of privacy, there remains a bigger question that even if the data leads to more targeted advertising, these measures taken by tech companies like Google or AI algorithms that detect depression via people’s social media feeds can lead to more accurate diagnoses, provide information about mental illness, lead to more cases being detected and more people being treated, so they may on balance be a good thing.

However, I am still very sceptical. To begin with, I doubt the extent to which these measures will have a positive impact. There is evidence, as in a 2015 study by Robert Epstein and Ronald Robertson published in the PNAS, to suggest that higher ranking and visibility on search results will influence people’s choices, so seeing the PHQ-9 on top of the results may encourage people to take the questionnaire and seek support. But that is hardly much of an improvement: at least in the UK, the top results when one I search for depression (using a new instance of private browsing to avoid personalised results as much as possible) are from the NHS, the mental health charity Mind, the Royal College of Psychiatrists and the Mental Health Foundation. Each of these links provides a wealth of resources about depression and ways to tell whether patients are showing signs of depression (including the PHQ-9). These are often more readily accessible and quicker to read than a 9-question survey. So really, the solution that Google promises is barely better than what is already available.

But the important thing here is context: while questionnaires like the PHQ-9 are useful in diagnosing depression, their accuracy and reliability have been debate. They are inadequate by themselves and they usually serve as just a preliminary indicator, often followed by a closer examination by a GP or a mental health professional. The trouble with Googling illness, however, is without the appropriate context of a GP or mental health professional to interpret results, presenting someone with a questionnaire can be dangerous. The phenomenon is called ‘cyberchondira’: confirmation bias while self-diagnosing symptoms and the lack of clear perspective can lead to diagnoses being exaggerated and misrepresented. This can cause people distress and anxiety. So taking this survey might not mean a patient is better informed, rather it might make them feel more anxious or distressed. This is my biggest issue with Google or Instagram being used to diagnose depression: the medium of diagnosis itself contributes to a patient’s distress or low sense of self-worth, et cetera. That this data can then be used for advertising — to take advantage of people when they are vulnerable — further exacerbates this.

I am wary of an emerging narrative right now, in the wake of Facebook and Google attempting to mitigate so-called ‘fake news’, that these tech giants are ‘taking responsibility’ for the information they provide, even though it is not really their job to do so. I am suspicious of this narrative because these companies, their impact on advertising, and their proprietary algorithms that filter search results and news feeds to reflect a user’s own views are part of the problem: they created the echo chambers that let fraudulent or partisan opinions to go unchallenged in the first place, and pay-per-click advertising made fake news such a lucrative industry. Besides, I found it odd that the first people to cover the story about Google were mostly business magazines or web sites (Forbes, the FT, Fortune and CNN Money) or tech magazines (Endgadget or Slashgear). It hardly made a positive impact in psychiatry or medicine. It is rather revealing that the story is predominantly entangled within the nexus of business and the tech industry, and those are the only stakeholders who view this in such positive light. For the most part, I doubt that this would yield any appreciable benefit.

Note: For resources on depression, including diagnoses and ways of living with chronic mental illness, see the NHS web site’s portal on clinical depression.