Journal Club: Rethinking Gray-Asexuality

Ace Journal Club banner

This month, the ace journal club discussed

“A Rethinking of Gray Asexuality: What do we Learn from an Undefinable Identity”, by Jason Gurevitch (2019). (freely available)

The journal club meets once a month on Discord, using text or voice as club members prefer. We discuss a variety of academic works in ace studies, ranging from gender studies to psychology. Don’t worry about journal access, we can provide access. If you’re interested, please e-mail me at for an invite.

Our discussion notes are below the fold.

The author explores the meaning of gray-asexuality by reviewing the literature, applying machine learning to the 2015 Ace Community Survey, and analyzing discussion in the AVEN forums. The analysis suggests that willingness to have sex is the most important factor distinguishing gray-asexuals, although not determinative. They argue that gray-asexuality is constructed through dual disidentification with both allosexuality and asexuality.

– This is an undergraduate honors thesis written by someone who double majored in women’s, gender, and sexuality studies, and computer science.
– Machine learning is a very unusual kind of analysis for this field. It may be difficult to get this kind of work peer-reviewed, as many reviewers may not understand it.
– As discussed in the thesis, the allosexual respondents to the Ace Community Survey still tend to have a connection to the ace community, and thus aren’t representative. I explained how the Ace Community Survey reports stopped emphasizing comparisons between ace and allosexual respondents for that reason.

– They used k-means clustering to identify groups of similar people in the Ace Community Survey.
– Initially, the clustering identified a group of people with more mental health problems. This is because the clustering algorithm assigns weight to each topic in proportion to the number of questions, and the Ace Community Survey had a lot of questions on that topic in 2015.
– After some of the questions were filtered out, three clusters were found: allosexuals, aces who have had sex, and aces who haven’t had sex.

Neural network
– They trained a neural network to classify people in the Ace Community Survey as asexual, gray/demi, or allosexual. The neural network was able to identify people who were more likely to be gray/demi, but it only had recall and precision around 50%.
– This suggests that there are group differences, but you can’t really distinguish groups on the individual level without asking people their identity.
– Willingness to have sex was the most important feature in distinguishing asexuals, gray/demis, and allosexuals.
– The neural network is similar in concept to the AIS developed by Yule et al. The Ace Community Survey found that gray, demi, and questioning people tend to fall in between asexual and allosexual people on the AIS, with significant overlap (see the 2017/2018 report, page 65).

Understanding gray-asexuality
– At one point, the author describes gray-asexuality as a “disorientation”. We thought this was an evocative term, suggesting that it’s like an orientation, but also a lack of orientation, a cognitive state of being in confusion. Although, it was also pointed out that gray-asexuals are not necessarily “confused” about being gray-asexual
– We liked the idea of gray-asexuality as double disidentification. I discussed my experience of feeling like I could identify as asexual, but choosing not to. Definitions are usually broad enough to include anyone who might get value of that identity, but even if I fit the definitions I might feel I get more value out of disidentifying with it.
– The thesis contrasts gray-asexuality with more specific labels like fraysexuality, saying that the specific labels deserve attention but that gray-asexuality is on a different level. This echoes my experience following google alerts, where there are whole articles about fraysexuality but none on gray-asexuality.
– We talked about why gray-asexuality gets so little attention. Gray-asexuality lacks any archetypical narrative, in contrast to asexuality. We talked about how asexual narratives follow a Pilgrim’s Progress narrative, which is about a journey of self-realization. A gray-asexual narrative would likely involve two self-realizations, and happen mostly internally.

– The author describes the common definition of gray-asexuality as not normally experiencing sexuality attraction, but experiencing it sometimes. The author frames this definition as flawed, and yet the best available. This definition appears to come from the AVENwiki, but it omits any of the other meanings that the AVENwiki lists directly below that one.
– The neural network’s finding about the importance of willingness to have sex doesn’t clearly fit into the central thesis.
– The neural network, which had 10 hidden layers and several thousand neurons seems a bit excessive considering the size of the data set.
– The thesis refers to “AceID”, which is an internally-used variable name in the Ace Community Survey. This is likely confusing to most would-be readers.
– The paper claims that gray-asexuality has been in use since AVEN was founded, but it was actually coined in 2006.

About Siggy

Siggy is an ace activist based in the U.S. He is gay gray-A, and has a Ph.D. in physics. He has another blog where he also talks about math, philosophy, godlessness, and social criticism. His other hobbies include board games and origami.
This entry was posted in Articles, Gray-A, Research and tagged . Bookmark the permalink.

2 Responses to Journal Club: Rethinking Gray-Asexuality

  1. ettinacat says:

    As a sex-repulsed grey-ace, I’m not exactly happy about that study’s conclusion.

    • Siggy says:

      Which part are you thinking about? Is it the neural network finding that willingness to have sex was the most important feature?

      In my interpretation, that just means that willingness to have sex was one of the most common reasons for people to disidentify with asexuality, but it doesn’t mean there are no other reasons. And there’s no reason for us to think that everyone ought to have the same reason.

      (Also, as a professional data scientist, I tend to be skeptical of any interpretation of “feature importance” unless I understand how exactly that’s defined. There are multiple ways to define it and the differences are significant.)

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s