What makes something a cult? Here is what our data says
- Travis M. (based on data analysis by Hudson Silvério)
- 21 minutes ago
- 12 min read
Short of time? Read the key takeaways.
💡 People rely on relational dynamics to judge cultishness. When people rate groups as cult-like, they focus most on signs of control, isolation, and suppression of outside information rather than on unusual predictions or strange practices.
🔍 Discouraging outside information is the strongest red flag. The clearest single predictor of perceived cultishness is when a group tries to limit members’ access to alternative viewpoints. People appear highly sensitive to this form of informational control.
🧠 Conformity pressure uniquely shapes perceptions. When considering all traits together, encouraging conformity is the most powerful predictor. This suggests that people see enforced sameness and behavioral control as central to what makes a group feel cult-like.
📉 Weirdness matters less than you might think. Uncommon rituals or unusual predictions correlate far less with perceived cultishness. People seem to distinguish harmless eccentricity from behaviors that restrict autonomy or isolate members.
📈 Our tool aligns closely with users’ intuition. Most assessment questions show large correlations with perceived cultishness, and the full model explains 63% of judgments. This suggests strong cult-literacy among users and supports the validity of our Cult Assessment tool.
You might have heard of academia referred to as a cult. After all, it demands personal sacrifice and comes with its own strange jargon, rigid hierarchies, and esoteric knowledge. Personally, I loved my time in academia, but I certainly saw plenty of my coworkers feeling overworked, undervalued, and struggling with imposter syndrome. This once led a colleague of mine to make the following poignant joke:
You know how I know academia isn’t a cult? Cults make you feel welcome.
It’s a funny joke, but would you say it’s also technically correct? Well, that depends on what you think makes something a cult. That’s partly what this article is about. In what follows, we’re going to explore some empirical findings from a recent study we ran, to answer questions like:
Which beliefs about groups tell us the most about how cult-like people think they are?
Which groups are people curious to assess as cults?
Read on to learn the answers to these questions and more!
What makes something a cult?
So what does make something a cult? There has been a lot of scholarship on this topic, and we synthesized some of that research (along with a study of our own) to build a tool that lets people assess how cult-like any given group is.
Our free online Cult Assessment tool asks you a question about each of our 15 criteria (listed below) for being a cult and generates a ‘Cult Score’ from your answers. To score highly, a group doesn't have to score highly on every single criterion, but the more highly it scores on more criteria, the higher the cult score it is assigned.
The Peoples Temple (which famously ended in a mass suicide/murder, the origin of the phrase “don't drink the kool-aid”) scores close to the maximum possible score of 100 since it’s high on almost every criterion. But many other very cult-like groups will still have some criteria that don't apply. The evaluation criteria we use are:
Clearer Thinking’s 15 Cult Criteria
Information is withheld from new members
Some types of external information seeking are discouraged
Confusions (e.g., about the group's actions or beliefs) don't get fully resolved
The group or its leader believes they uniquely possess important knowledge
The group views itself as special
The group members share uncommon predictions about the future
The group has non-typical practices
Large monetary contributions are expected
Large contributions of time are expected
Self-sacrifice is expected
Contact with former members is discouraged, or former members are deeply judged
Close ties with non-members are discouraged
The group is viewed negatively by unconnected non-members who know about its activities
The group discourages members from being individualistic or from being different from others in the group
The group engages in practices that would be considered unethical by many in the surrounding culture
And the questions we ask about each criterion work like this: Users are shown a spectrum and asked to consider where the group they’ve chosen “falls on the spectrum between the top item (labeled ‘0’) and the bottom item (labeled ‘4’).” They are instructed:
If the group is just like the top item, then select the top item (labeled '0').
If it is somewhat more like the top item than the bottom item, choose the option labeled '1'.
If it is halfway between the top and bottom item, choose 2.
If it is somewhat more like the bottom item than the top item, choose 3.
If it is just like the bottom item, choose 4.
An example spectrum is shown here:
Some types of external information seeking are discouraged
0. Group members are not at all discouraged from seeking outside information about the group (that’s not created by current group members), and from seeking outside perspectives or information sources that disagree with the group’s beliefs. |
1. |
2. |
3. |
4. Group members are forbidden or strongly discouraged from viewing information about the group made by people not currently in the group, or from seeking outside perspectives or information sources that contradict the group’s beliefs. |
A higher cult score indicates a higher chance that the group in question is a harmful cult. Though, of course, a high score does not guarantee it is a harmful cult, and there will inevitably be some harmful groups that don't get a high score. To learn more about how we settled on these criteria, you can read the optional “How this tool was developed” section of our free online Cult Assessment tool.
Roughly nine months after we launched our tool, we took a look at the anonymized, non-sensitive data that 1528 people who completed the tool provided, to see what we could learn. Below are some of our initial findings.
What do people see in groups they think are cults?
In addition to calculating a Cult Score (based on your answers to 15 questions about cult-like traits), our tool asks you other questions, including:
What is your relationship to this group?
(For example: Current member, former member, etc.)
And, after all other questions:
On a scale from 0% (not at all a cult) to 100% (wholly a cult), where would you rate this group, community or organization?
Let's call this "Perceived Cultishness".
We calculated the correlations between Perceived Cultishness scores and each of the 15 assessment questions, for all (and only) people who said their relationship to the group they were assessing was either: “I am a current group member” or “I am a former group member”. By doing this, we were able to gain some insight into what people see in groups they think are cults. The graph below shows the results:
To put it simply: These correlations tell you how useful each question would be if you had to predict how cult-like people think a group is, but were allowed to ask only one question. The stronger the correlation, the better it is for predicting.
The fact that so many of the correlations exceed 0.5 is quite remarkable: In the field of psychology, it's more common to get correlations with magnitudes in the 0.2 to 0.4 range, so correlations of 0.5 or 0.6 are notably large! All-but-three of our correlations count as large (by standard conventions in social science), while two are medium-large, and the smallest correlation is still medium. The strengths of these correlations tell us that the traits we're asking about are tightly linked to people's intuitive perception of what is a cult.
Of course, the correlations by themselves don’t reveal whether the traits influence the judgments that people make, or their judgments (of how cult-like a group is) influence the trait ratings, or some third factor influences both.
Given that most of our assessment questions were derived from existing research and literature on cults, these results also suggest there may be a fairly high level of cult-literacy among people who have used our tool. That’s good to see!
Another finding is that the strongest individual predictor of Perceived Cultishness (in our sample) is discouraging the seeking out of external information. And we agree that when a group discourages you from seeking information from non-group sources, that's a worrisome sign.
In our article about how to spot real expertise, we wrote about how people with genuine expertise will often point to sources beyond themselves. As we said there:
Genuine experts don't seek to be seen as a sole arbiter of knowledge or authority on a topic (which can be an indication that ego, rather than truth-seeking, is a primary motivation for them), but instead encourage you to look at resources other than the ones they have produced.
This is because genuine expertise comes with the knowledge that you are only one node in a giant web of knowledge production. Perhaps the most famous expression of this fact comes from a letter Isaac Newton once wrote to fellow scientist Robert Hooke, wherein Newton responded to praise of his genius by writing:
[Y]ou defer too much to my ability in searching into this subject. What Descartes did was a good step. You have added much several ways, and especially in considering the colours of thin plates. If I have seen farther, it is by standing on the shoulders of giants.
It's heartening that our results suggest people are wary of groups that discourage them from seeking external information.
The last finding we want to talk about from the individual correlations is that the weakest individual predictor of Perceived Cultishness (in our sample) is making unusual predictions (though it still had a meaningful positive correlation of r=0.36). Similarly, unusual practices (which fits a similar theme) was second-weakest (r=0.41).
This suggests that people’s conceptions of cultishness are not primarily about fringe predictions or prophecies. Instead, by looking at the top predictors and comparing them to the bottom predictors, we see some evidence that people’s conceptions of cultishness are more essentially built around relational and behavioral dynamics that involve control, isolation, and pressure. This is consistent with the existing cult assessment frameworks we studied in the literature (see the “Other Assessment Frameworks” section of our free online Cult Assessment tool for more details) and so adds to the impression that users of our tool appear to have a fairly high level of cult-literacy.
Which beliefs about groups are the best predictors of how cult-like they seem?
In addition to calculating individual correlations between each of our 15 assessment questions and the Perceived Cultishness, we also ran a multivariate linear regression to find out how much unique work each assessment question does at predicting Perceived Cultishness above and beyond (i.e., when controlling for or accounting for) all the other ratings. For this analysis, we also controlled for whether the person using the tool was a current or former member of the group they were assessing. Here is a graph showing the results:
The individual correlations in the previous section told us how predictive each assessment question is by itself, whereas these results tell us how predictive each assessment question is when already taking all of the others into account. The individual analysis is useful for understanding the effects of single variables. A possible use case for this prior analysis (which relies on a causal connection) could be if you wanted traction on the question:
If I want to convince people that this group is a cult in a short amount of time, which single trait could it be most effective to focus on?
The regression analysis is useful when you want to understand how the perceptions of traits work together. Such as if you wanted to answer the question:
If I want to predict how cultish people think a group is, which group of traits are most useful for making that prediction?
The value of R² for this analysis is 0.63, meaning that 63% of the variation in Perceived Cultishness is explained by the prediction made from all the assessment questions. And the ‘Current members’ bar is included because we adjusted for whether each participant was a current member or former member of the group they were rating. It shows that when controlling for all of the ratings on the individual attributes of the group, current members (unsurprisingly) are less likely to rate groups as cults.
Optional note on how to read regression coefficients To understand the regression coefficients we use in this article, you need some understanding of the notion of ‘standard deviation’, so let’s start there. Standard deviation is a way of quantifying how spread out data is. Roughly speaking, the standard deviation is the ‘typical’ (average) amount of distance your data points are from the mean value in the data. More technically, it's the square root of the sum of squares of distances to the mean, but if that sounds too complicated, you can just think of it as the typical distance from the mean. Now, a regression coefficient of 0.14 (such as for ‘Conformity encouraged’) means that when this rating increases by 1x its standard deviation, Perceived Cultishness also increases by 0.14x its standard deviation, on average, holding other assessment questions constant. Because our assessment questions are conceptually related (they’re all cult characteristics) we can expect that people’s ratings of them will move together somewhat (i.e., they'll be correlated). This is called ‘multicollinearity,’ and it tends to shrink each coefficient because the coefficient gets spread out across the (similar) factors. Even though these coefficients may look modest, the regression model as a whole does a good job of predicting how cult like people say that groups are (as reflected by the R2 value of 0.63, meaning that 63% of the variation in cultishness ratings are captured by these variables). |
The trait that most powerfully predicts Perceived Cultishness, when taking into account the effects of all the other traits, is the extent to which conformity is encouraged.
In the individual correlation analysis, the most predictive trait was discouraging seeking information from external sources but, in the regression analysis, encouraging conformity has jumped to the top. One possible explanation for this is that, although both are important in people's assessments, the extent to which conformity is encouraged was a more unique criterion that was not already captured in the other questions as much, and so controlling for the other effects doesn't reduce its usefulness as much as for the trait of discouraging external information seeking.
It is interesting to observe that the three traits most predictive of Perceived Cultishness in the regression analysis all relate to isolating members in a homogeneous group. None of the least predictive traits relates to this theme.
It's also interesting that in both analyses (correlation and regression), high levels of sacrifice (of money, time, or self) are in the middle or bottom of the pack of predictors of Perceived Cultishness. This may surprise some people because part of what makes cults so harmful is that they extract resources from their members, including their time (e.g., doing work on behalf of the group) and money. While that’s true, perhaps these factors aren't as strong predictors as others because they are less unique to cults. For instance, if you’ve ever been very engaged in an expensive hobby or competed in something at a high level, you’ve probably experienced some or all of these demands for sacrifice. There are also lots of jobs and non-profit organizations that ask for a great deal of people’s time.
As with the correlation analysis, we see in this analysis too that, that unusualness (in practices and predictions) is the weakest kind of predictor. So, what you might think of as ‘weirdness’ is not very predictive of Perceived Cultishness, either by itself or when taking into account other factors.
When controlling for the other questions, neither making uncommon predictions about the future nor requiring large contributions of time has a statistically significant effect on Perceived Cultishness. This might be because their effects were already taken into account by other questions in the set - so they may not be needed for evaluating Perceived Cultishness when the other questions are already being asked.
Finally, out of curiosity, we also looked into which groups were assessed more than once by users of our tool. The results are presented in this graph:
Of course, these numbers are most determined by a combination of (a) who our readership is, and (b) what groups our readership is interested in evaluating. A group’s presence in this chart does not entail that it is a cult.
Although no robust conclusions can be inferred from the results depicted in this graph, we find it interesting to note that groups being assessed come from a variety of areas of life (religious, social, political, self-help, and beyond) and that some people using our tool thought that groups like family and personal work were worth assessing for cult-like traits.
What does all this mean?
So, what can you take away from these findings? Our data provide reasons to be optimistic: People’s judgments about the cultishness of groups seem to match well with the literature and with our tool’s criteria for assessment. This suggests that cult-literacy is reasonably high, at least among our sample. However, since we performed our analyses on the results from people who voluntarily chose to use a tool about cult assessment and reported being either current or former members of the groups they were assessing, we might expect their judgments to be better than the average person’s.
These findings also provide some validation for our cult assessment tool. Since we observed large correlations for most of our assessment questions and an R² of 0.63 when predicting Perceived Cultishness, we can infer some degree of convergent validation (i.e., what we use to evaluate cultishness strongly matches the way that current and former members make evaluations). We also observed consistency with what we expected, based on theory. These are reasons to increase confidence in the effectiveness of our tool.
If you want to try it out for yourself, or learn more about how it was developed, you can do so right now (it's free):