Conference call, 9 December 2009

Previously sex prediction work

74% accuracy predicting offline sex most of the respondents were females playing as female

problem: males playing as females identified as females

maybe behavior or social network data
looking at places where "males and females behave differently independent of their avatar sex/gender"

Bassiean (sp) model in machine learning technique

intro 80s, popular recently
allows you to specify qualitative causal relationships
- e.g. females tend to be more social than males, "frequency of chat"

Need variables, measures

What differentiates men + women in general?
SL: "number of chars they use to chat"
Social networks, people you work with
- RW lit: men have larger nets with weaker ties, women have smaller nets w stronger ties
Money
- How much they buy?
- Spend?
- Make?
Identity
- Time spent making onself up
- Money spent on appearance related items (not so much in EQ2)
Self-presentation
- Women with their boobs out?
- Body construction?
- Clothing choice
Living
- Clustering geographically
Using voice chat
- Consistent with their avatar/offline
- When cross-av dressing

SL v EQ2

If the model doesn't fit the RW expectations, is the model broken? Or might the population of VW/MMO users not reflect the RW expectations?

If so, it may be an effect of VW/MMO participation or a pre-existing preference for spaces with different sex/gender expectations?

Strategic v performative gender variance

Machine learning

Social net