Psychedelic Drug Research and the Data-Mining Revolution
One of the most mysterious problems in neuroscience is the link between brain chemistry and consciousness. How do changes in our neurochemistry influence our perception of the real world?
This question is hard to tackle for the obvious reason that experiments on humans are notoriously difficult to perform. Not only are the variables hard to pin down but changing them with psychoactive drugs under controlled conditions is fraught with practical, ethical, and moral dilemmas.
That’s why the majority of work examining the role of psychoactive drugs on neuropharmacological signaling mechanisms has been done on rats.
But there’s a revolution afoot. Today, Jeremy Coyle at the University of California Berkeley and a couple of pals say they’ve found a new way to study the role of psychoactive drugs on human perception.
These guys point out that in the contrast to the small amount of formal scientific literature in this area, there are large volumes of narrative descriptions of the effects of drugs posted on the web. Their idea is to mine these descriptions using machine learning techniques to identify common features which would allow a quantitative comparison of their effects.
The obvious place to start such an endeavour is a website called erowid.org, which is a well known and popular source of user generated information about the effects of all kinds of psychoactive substances.
Coyle and co confine their investigations to ten drugs ranging from
3,4‐methylenedioxymethamphetamine, better known as ecstacy, and lysergic acid diethylamide, or LSD, to less well known drugs such as N,N‐dipropyltryptamine, sometimes called The Light, and
2,5‐dimethoxy‐4‐ethylphenethylamine which has the street name Europa.
They collected 1000 narrative reports on these drugs and mined the text for common words, while screening out some words that are common to more than five drugs.
Having identified signature words, they then tested their hypothesis by seeing whether the results could be used to accurately predict which drug the reports referred to.
It turns out that some drug reports are much easier to classify than others. Ecstasy reports tends to use words such as “club”, “hug”, “rub” and “smile”, which reflect the social setting in which the drug is often used and the feelings of love and friendliness the drug seems to produce.
By looking for clusters of these words, Coyle and co were able to accurately identify Ecstasy reports almost 90 per cent of the time, by far the highest success rate in this experiment.
On average, they were able to correctly classify the drug reports about 50 per cent of the time, which is significantly better than the 10 per cent success (1 in 10) expected by chance.
Coyle and co were also able to identify clusters of words that were common between different drugs. For example, LSD and magic mushrooms are both associated with words such as “see”, “look”, “saw”, “room”, “tell”, “ask”, “walk”, “house”.
And drugs called DMT and Salvia are associated with words such as “reality”, “dimension”, “universe”, “state”, “consciousness”, “form”, “entity”. “Both drugs have been associated with powerful alterations in consciousness and feelings of altered reality,” say Coyle and co.
What’s interesting about these clusters of words is that it allows Coyle and co to speculate about the receptors and signalling pathways in the brain that the drugs target.
For example, the lesser known drugs The Light and Europa were associated with words such as “stomach”, “nausea”, “vomit”, “headache”.
Coyle and co say that stimulation of the 5‐HT-3 receptor is thought to induce nausea and vomiting, which leads to a straightforward hypothesis: “These two substances, more than the other studied psychedelics, may directly stimulate 5‐HT-3 receptors or may induce release of 5‐HT from enterochromaffin cells,” they say.
That’s an interesting and significant step forward: the potential identification of the receptors and signalling pathways involved in the conscious perception of reality.
Of course, there are limitations to this kind of approach. Coyle and co cannot know how many of the reports are inaccurate or mislabelled.
Also, differences in drug narratives are likely to reflect the age and sex of the users as well as the social circumstances in which the drugs were used. However, Coyle and co say that it may be possible to control for these variables in future studies if more user data becomes available.
Given the limitations on human experimentation in this area, Coyle and co have found a novel way of teasing apart the neurochemistry of consciousness using data mining and machine learning techniques. Interesting stuff!
Ref: arxiv.org/abs/1206.0312: Quantitative Analysis of Narrative Reports of Psychedelic Drugs
The inside story of how ChatGPT was built from the people who made it
Exclusive conversations that take us behind the scenes of a cultural phenomenon.
ChatGPT is about to revolutionize the economy. We need to decide what that looks like.
New large language models will transform many jobs. Whether they will lead to widespread prosperity or not is up to us.
Sam Altman invested $180 million into a company trying to delay death
Can anti-aging breakthroughs add 10 healthy years to the human life span? The CEO of OpenAI is paying to find out.
GPT-4 is bigger and better than ChatGPT—but OpenAI won’t say why
We got a first look at the much-anticipated big new language model from OpenAI. But this time how it works is even more deeply under wraps.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.