Problems with Bargh’s definition of unconscious

iceberg_cutI have a new paper out in Frontiers in Psychology: The perspectival shift: how experiments on unconscious processing don’t justify the claims made for them. There has been ongoing consternation about the reliability of some psychology research, particularly studies which make claims about unconscious (social) priming. However, even if we assume that the empirical results are reliable, the question remains whether the claims made for the power of the unconscious make any sense. I argue that they often don’t.

Here’s something from the intro:

In this commentary I draw attention to certain limitations on the inferences which can be drawn about participant’s awareness from the experimental methods which are routine in social priming research. Specifically, I argue that (1) a widely employed definition of unconscious processing, promoted by John Bargh is incoherent (2) many experiments involve a perspectival sleight of hand taking factors identified from comparison of average group performance and inappropriately ascribing them to the reasoning of individual participants.

The problem, I claim, is that many studies on ‘unconscious processing’, follow John Bargh in defining unconscious as meaning “not reported at the time”. This means that experimenters over-diagnose unconscious influence, when the possibility remains that participants were completely conscious of the influence of the stimili, but may not be reporting them because they have forgotten, worry about sounding silly or because the importance of the stimuli is genuinely trivial compared to other factors.

It is this last point which makes up the ‘perspectival shift’ of the title. Experiments on social priming usually work by comparing some measure (e.g. walking speed or reaction time) across two groups. My argument is that the factors which make up the total behaviour for each individual will be many and various. The single factor which the experimenter is interested in may have a non-zero effect, yet can still justifiably escape report by the majority of participants. To make this point concrete: if I ask you to judge how likeable someone is on the 1 to 7 scale, your judgement will be influenced by many factors, such as if they are like you, if you are in a good mood, the content of your interaction with the person, if they really are likeable and so on. Can we really expect participants to report an effect due to something that only the experimenter sees variation in, such as whether they are holding a hot drink or a cold drink at the time of judgement? We might as well expect them to report the effect due to them growing up in Europe rather than Asia, or being born in 1988 not 1938 (both surely non-zero effects in my hypothetical experiment).

More on this argument, and what I think it means, in the paper:

Stafford, T. (2014) The perspectival shift: how experiments on unconscious processing don’t justify the claims made for them. Frontiers in Psychology, 5, 1067. doi:10.3389/fpsyg.2014.01067

I originally started writing this commentary as a response to this paper by Julie Huang and John Bargh, which I believe is severely careless with the language it uses to discuss unconscious processing (and so a good example of the conceptual trouble you can get into if you start believing the hype around social priming).

Full disclosure: I am funded by the Leverhulme Trust to work on a project looking at the philosophy and psychology of implicit bias. This post is cross-posted on the project blog.

Seeing ourselves through the eyes of the machine

I’ve got an article in The Observer about how our inventions have profoundly shaped how we view ourselves because we’ve traditionally looked to technology for metaphors of human nature.

We tend to think that we understand ourselves and then create technologies to take advantage of that new knowledge but it usually happens the other way round – we invent something new and then use that as a metaphor to explain the mind and brain.

As history has moved on, the mind has been variously explained in terms of a wax tablets, a house with many rooms, pressures and fluids, phonograph recordings, telegraph signalling, and computing.

The idea that these are metaphors sometimes gets lost which, in some ways, is quite worrying.

It could be that we’ve reached “the end of history” as far as neuroscience goes and that everything we’ll ever say about the brain will be based on our current “brain as calculation” metaphors. But if this is not the case, there is a danger that we’ll sideline aspects of human nature that don’t easily fit the concept. Our subjective experience, emotions and the constantly varying awareness of our own minds have traditionally been much harder to understand as forms of “information processing”. Importantly, these aspects of mental life are exactly where things tend to go awry in mental illness, and it may be that our main approach for understanding the mind and brain is insufficient for tackling problems such as depression and psychosis. It could be we simply need more time with our current concepts, but history might show us that our destiny lies in another metaphor, perhaps from a future technology.

I mention Douwe Draaisma’s book Metaphors of Memory in the article but I also really recommend Alison Winter’s book Memory: Fragments of a Modern History which also covers the fascinating interaction between technological developments and how we understand ourselves.

You can read my full article at the link below.

Link to article in The Observer.

Awaiting a theory of neural weather

In a recent New York Times editorial, psychologist Gary Marcus noted that neuroscience is still awaiting a ‘bridging’ theory that elegantly connects neuroscience with psychology.

This reflects a common belief in cognitive science that there is a ‘missing law’ to be discovered that will tell us how mind and brain are linked – but it is quite possible there just isn’t one to be discovered.

Marcus, not arguing for the theory himself, describes it when he writes:

What we are really looking for is a bridge, some way of connecting two separate scientific languages — those of neuroscience and psychology.

Such bridges don’t come easily or often, maybe once in a generation, but when they do arrive, they can change everything. An example is the discovery of DNA, which allowed us to understand how genetic information could be represented and replicated in a physical structure. In one stroke, this bridge transformed biology from a mystery — in which the physical basis of life was almost entirely unknown — into a tractable if challenging set of problems, such as sequencing genes, working out the proteins that they encode and discerning the circumstances that govern their distribution in the body.

Neuroscience awaits a similar breakthrough. We know that there must be some lawful relation between assemblies of neurons and the elements of thought, but we are currently at a loss to describe those laws.

The idea of a DNA-like missing component that will allow us to connect theories of psychology and neuroscience is an attractive one, but it is equally as likely that the connection between mind and brain is more like the relationship between molecular interactions and the weather.

In this case, there is no ‘special theory’ that connects weather to molecules because different atmospheric phenomena are understood in multiple ways and across multiple models, each of which has a differing relationship to the scale at which the physical data is understood – fluid flows, as statistical models, atomic interactions and so on.

In explanatory terms, ‘psychology’ is probably a lot like the weather. The idea of their being a ‘psychological level’ is a human concept and its conceptual components won’t neatly relate to neural function in a uniform way.

Some functions will have much more direct relationships – like basic sensory information and its representation in the brain’s ‘sensotopic maps’. A good example might be how visual information in space is represented in an equivalent retinotopic map in the brain.

Other functions will have more more indirect relationships but in great part because of how we define ‘functions’. Some have very empirical definitions – take iconic memory – whereas others will be cultural or folk concepts – think vicarious embarrassment or nostalgia.

So it’s unlikely we’re going to find an all-purpose theoretical bridge to connect psychology and neuroscience. Instead, we’ll probably end up with what Kenneth Kendler calls ‘patchy reductionism’ – making pragmatic links between mind and brain where possible using a variety of theories and descriptions.

A search for a general ‘bridging theory’ may be a fruitless one.

Link to NYT piece ‘The Trouble With Brain Science’.

Towards a scientifically unified therapy

nature_scienceToday’s edition of Nature has an excellent article on the need to apply cognitive science to understanding how psychological therapies work.

Psychological therapies are often called ‘talking treatments’ but this is often a misleading name. Talking is essential, but it’s not where most of the change happens.

Like seeing a personal trainer in the gym, communication is key, but it’s the exercise which accounts for the changes.

In the same way, psychological therapy is only as effective as the experience of putting changes into practice, but we still know relatively little about the cognitive science behind this process.

Unfortunately, there is a traditional but unhelpful divide in psychology where some don’t see any sort of emotional problem as biological in any way, and the contrasting divide in psychiatry where biology is considered the only explanation in town.

The article in Nature argues that this is pointless and counter-productive:

It is time to use science to advance the psychological, not just the pharmaceutical, treatment of those with mental-health problems. Great strides can and must be made by focusing on concerns that are common to fields from psychology, psychiatry and pharmacology to genetics and molecular biology, neurology, neuroscience, cognitive and social sciences, computer science, and mathematics. Molecular and theoretical scientists need to engage with the challenges that face the clinical scientists who develop and deliver psychological treatments, and who evaluate their outcomes. And clinicians need to get involved in experimental science. Patients, mental-health-care providers and researchers of all stripes stand to benefit.

The piece tackles many good examples of why this is the case and sets out three steps for bridging the divide.

Essential reading.

Link to ‘Psychological treatments: A call for mental-health science’.

The concept of stress, sponsored by Big Tobacco

NPR has an excellent piece on how the scientific concept of stress was massively promoted by tobacco companies who wanted an angle to market ‘relaxing’ cigarettes and a way for them to argue that it was stress, not cigarettes, that was to blame for heart disease and cancer.

They did this by funding, guiding and editing the work of renowned physiologist Hans Selye who essentially founded the modern concept of stress and whose links with Big Tobacco have been largely unknown.

For the past decade or so, [Public Health Professor Mark] Petticrew and a group of colleagues in London have been searching through millions of documents from the tobacco industry that were archived online in the late ’90s as part of a legal settlement with tobacco companies.

What they’ve discovered is that both Selye’s work and much of the work around Type A personality were profoundly influenced by cigarette manufacturers. They were interested in promoting the concept of stress because it allowed them to argue that it was stress — not cigarettes — that was to blame for heart disease and cancer.

“In the case of Selye they vetted … the content of the paper, they agreed the wording of papers,” says Petticrew, “tobacco industry lawyers actually influenced the content of his writings, they suggested to him things that he should comment on.”

They also, Petticrew says, spent a huge amount of money funding his research. All of this is significant, Petticrew says, because Selye’s influence over our ideas about stress are hard to overstate. It wasn’t just that Selye came up with the concept, but in his time he was a tremendously respected figure.

Despite the success of the campaign to associate smoking with stress relief, the idea that smoking alleviates anxiety is almost certainly wrong. It tends to just relieve anxiety-provoking withdrawal and quitting smoking reduces overall anxiety levels.

Although the NPR article focuses on Selye and his work on stress, another big name was recruited by Big Tobacco to promote their theories.

It’s still little known that psychologist Hans Eysenck took significant sums of cash from tobacco companies.

They paid for a lot of Eysenck’s research that tried to show that the relationship between lung cancer and smoking was not direct but was mediated by personality differences. There was also lots of other research arguing that a range of smoking related health problems were only present in certain personality types.

Tobacco companies wanted to fund this research to cite it in court cases where they were defending themselves against lung cancer sufferers. It was their personalities, rather than their 20-a-day habit, that was a key cause behind their imminent demise, they wanted to argue in court, and they needed ‘hard science’ to back it up. So they bought some.

However, the link between ‘father of stress’ Hans Seyle and psychologist Hans Eysenck was not just that they were funded by the same people.

A study by Petticrew uncovered documents showing that both Seyle and Eysenck appeared in a 1977 tobacco industry promotional film together where “the film’s message is quite clear without being obvious about it — a controversy exists concerning the etiologic role of cigarette smoking in cancer.”

The ‘false controversy’ PR tactic has now became solidified as a science-denier standard.

Link to The Secret History Behind The Science Of Stress from NPR.
Link to paper ‘Hans Selye and the Tobacco Industry’.

Motherhood, apple pie and replication

Who could possibly be against replication of research results? Jason Mitchell of Harvard University is, under some conditions, for reasons described in his essay On the emptiness of failed replications.

I wrote something for the Centre for Open Science which tries to draw out the sensible points in Mitchell’s essay – something I thought worth doing since for many people being against replication in science is like being against motherhood and apple pie. It’s worth noting that I was invited to do this by Brian Nosek, who is co-founder of the Center for Open Science and instrumental in the Many Labs projects. As such, Brian is implicitly one of the targets of Mitchell’s criticisms, so kudos to him for encouraging this discussion.

Here’s my commentary: What Jason Mitchell’s ‘On the emptiness of failed replications’ gets right

A spook’s guide to the psychology of deception

Last February, a file from the Edward Snowden leaks was released from a 2012 GCHQ presentation called ‘The Art of Deception: Training for Online Covert Operations’. It describes the ‘Online Covert Action Accreditation’ course which draws heavily on the psychology of influence and persuasion. This post will look at how they’re piecing together the science that forms the basis for these online operations.

The work seems to have been put together by GCHQ’s Human Science Operations Cell which presumably exists as an internal consultancy to allow the relevant cognitive and social sciences to be applied to practical covert operations.

One of the early slides lists the subjects the HSOC draws on which stretch from psychology to political science to neuroscience. At the current time, neuroscience has nothing practical to contribute, so they’re clearly blowing their neurological trumpets to sound a bit more high-tech but it’s worth noting the breadth of disciplines they draw on meaning they’ve got a wide and comprehensive vision of human behaviour from the micro to the macro.

However, one of the key slides has a road map of how everything fits together. It’s shown below and it’s quite dense so you can click the image below if you want a larger version.

One of the first thing that stands out is the ad-hoc-ness of their approach. They’ve appropriated a patchwork of relevant theories as a guide to practice with nothing being drawn from their own data.

You can see the main areas they’re drawing from – which includes profiling cultures and personality, research on persuasion, cognitive biases and scams, research on the psychology of stage magic, and organisational psychology or management science more generally.

Perhaps the weakest elements here are the cultural and personality profiling using Hofstede’s cultural dimensions and a Big Five personality traits. The trouble is that while these are statistically reliable on the group level they predict very little on the individual level because the effects are swamped by individual variation.

This means it may be more useful in the domain of PSYOPS, which attempts to influence groups, rather than targeting individuals.

The slide below details the general psychological framework for deception. As far as I can tell, this is the only original piece of psychological theory in the presentation.

It’s more a useful way of organising different approaches to deception rather than a theory in itself. It’s what clinical psychologists would call a ‘formulation’. It’s a way of organising evidence-based effects that may not be thoroughly tested itself but works well enough to aid understanding.

Perhaps the key thing to note is the sensemaking component. Sensemaking is a key concept in management science that just describes the different ways in which people come to conclusions about the meaning and significance of things.

It should be a well-known concept in intelligence circles because it is used both in military people management and military intelligence analysis. Interestingly, they treat individuals as like naive intelligence analysts who are trying to piece together their own understanding of the world and aim to exploit some of the weaknesses in this process. The big messy ‘concept map’ slides mentions ‘destructive organisational psychology’ which presumably refers to using the understanding of what keeps organisation together to break them apart.

However, in terms of the psychological science which underlies their approach, the next slide is key.

You can see several influences here. The techniques listed under ‘attention’ are all taken from research on the psychology of magic tricks, particularly from Susana Martinez-Conde’s work on how sleight-of-hand artists manipulate attention. Most of it is reviewed in a paper she wrote with a series of co-authors including pickpocket Apollo Robbins.

The HSOC spooks clearly love the idea of the psychology of magic and they refer to it a lot in their presentation. One slide just says ‘We want to build Cyber Magicians’, but it’s really not clear how it applies online. The whole point of sleight-of-hand is that it is dynamic and takes advantage of how you pay attention. When online, however, users’ attention doesn’t necessary flow in a predictable pattern because you can wander off from the screen, pause, grab screenshots and so on. In other words, individuals have better control over the flow of information because online interaction is designed for information control and therefore partial staggered attention.

The ‘perception’ techniques listed on the slide are largely taken from Stefano Grazioli and Sirkka Jarvenpaa’s classic paper [pdf] on online deception entitled ‘Deceived: Under Target Online’. The paper looks at how internet scammers rip people off and assuming that successful online con artists have found useful techniques by natural selection, HSOC just borrow them.

The techniques to exploit sensemaking are largely based on theories of sensemaking itself although the story fragments components seems to be drawn from research on relational agents that are designed ‘to form long-term, social-emotional relationships with their users’. Rather than actually deploying autonomous relational agents, I suspect it’s simply a case of using research insights from the area that suggests, for example, that presenting fragments of the agent’s backstory and letting the other person piece them together makes the person seem more believable.

The techniques in the ‘affect’ section are some general points taken from a vast experimental literature on the psychology of marketing and persuasion that describes how emotion modulates the heuristics (judgement processes) involved in persuasion.

The ‘behaviour’ section is the only part I don’t recognise as coming from the psychological literature. This makes me suspect it comes from PSYOPS or IO practice, but if you recognise it, leave a comment below.

The ’10 Principles of Influence’ is perhaps one of the most interesting slides in terms of illustrating the empirical basis for their approach as they use research both on the strategies of honest persuasion and dishonest scammers.

‘Principles are influence’ are largely associated with the work of consumer psychologist Robert Cialdini but the list actually consists of three of his six principles (Reciprocity, Social Compliance / Authority, Consistency).

Another six are taken from Stajano and Wilson’s classic study ‘Understanding scam victims: seven principles for systems security’ which describes six methods used by con artists. One item overlaps with the Cialdini principles and additionally they’ve included flattery (known to be an effective persuasive tool) and time – although it’s not clear whether they’re referring to giving people time and putting people under time pressure.

This section seems to be about gaining people’s trust to encourage disclosure and the slide you see above refers to social penetration theory which describes how relationships progress to increased levels of intimate connection through self-disclosure. The slide that follows this gives some basic advice about encouraging this: mirroring communication cues, adjust speech patters and so on – the sort of things you get taught in the first week of a psychotherapy course.

So here’s what the Online Covert Action Accreditation’ course looks like: like a PhD psychologist was given the task to come up with a plausible psychological framework for practical deception and influence online. It draws on a mix of persuasion psychology from marketing, studies on scammers and con-men, the social psychology of trust and disclosure, studies of how stage magic works psychologically, and work on what makes organisations work effectively and what degrades their performance.

This is a comprehensive approach to the problem, but the trouble is, this probably only translates approximately and probably rather poorly into practical effects.

In place of this, HSOC would be better of doing research and lots of it. They could do lots of informal RCTs online and gather a large amount of data quite quickly to test out which techniques actually increase influence or lead to successful deception. What behaviours on the part of the actor lead to increased self-disclosure the quickest? Does a laggy internet connection mean people’s increased frustration affects their evaluation of honest? and so on.

I suspect, however, that the Human Science Operations Cell were, and maybe still are, quite a small outfit and so they’re restricted to a consultancy role which will ultimately limit their effectiveness.

We tend to think that the secret services are super efficient experts with an infinite budget, but they probably just work like any other organisation. HSOC were probably told to deliver an Online Covert Action Accreditation course with few resources and not enough time and came up with the most sensible thing in the time allowed.

Oh, and by the way, hello spooks, and welcome to Mind Hacks.

Link to copy of slides.
Link to coverage from The Intercept.