At some point in the past, I noticed that I had a strong tendency to make NPCs male, even though there wasn’t any good story or setting-specific reason to do so. From gods to villains to random shopkeepers - most of these were assigned male without me even realizing that I have been doing it.
Thus, I started to assign genders by the roll of a dice - and I am fairly pleased with the results as this made the world significantly more diverse.
How about you? Have you noticed any similar biases in your own NPCs - and if so, what did you do about this?
I roll on a table for sex and gender when making random NPCs. If I don’t, I’ll design the entire character without ever assigning any of that and won’t realize it until playtime.
I did this in a novel I wrote, actually. I assigned TLA ‘names’ based on their job (ENG, PIL, etc), and any time a gender would normally be referenced in the text I used XXX - both for easy searching. I got about 70% of the way through when my beta readers rebelled - they absolutely HAD TO KNOW what gender everybody was. Sigh.
But by this time the characters’ personalities and speech patterns were well established, so I flipped a coin for each one, and continued onward. I’ll probably do this again some day and just ignore the beta readers.
Yes, when prepare NPC I always have dice determine gender/species/culture/etc to avoid my own biases. On the spot I atleast try to determine gender randomly.
Though I understand for gender, I think a lot of those depend a lot on the setting.
I’m genderfluid, I write whoever I wanna.
Yes. I use chat AI for this kind of thing, you can ask it to generate a table and then roll on the table, or just generate a list of NPCs and ask for a % representation of people.
How well does chat AI do when it comes to representation?
One huge flaw of AI-based systems I keep on hearing about is that they tend to repeat the biases of their training data.
If you ask for a random sample I feel like you get a fairly even distribution of male/female/lgbtq, but sometimes yes you might get a sample that’s all men or women or whatever. So yes it can be biased. It’s the luck of the draw.
The good thing is you can just tell it the exact proportions you want, or even get it to just give you more variety. If you say “Give me 50/50” or “give me a table of all possibilities, then base it off an equal distribution” or simply “Wow, that’s a lot of dudes/kind of stereotypical, can you give me more variety?” or even “This data set seems biased, can you give me suggestions for more variety”. If you talk in dnd terms and go like ‘1dx table with the roles gender/sex/race/orientation’ or whatever, it will do that, it’ll even show you the dice rolls if you demand it.
So initial impressions can be biased…but your own brain is biased too. The good thing is you can generate a whole host of npcs literally in the time it takes you to roll a die 10 times lol. You can even tell it to combine disciplines by doing something like to literally combat it’s own biases by adding in thoughts about DEI/representing other people if you want. Or you can go opposite and tell it to be ‘conservative’ if you wanna be like that.
Heck, you can even literally feed it what you posted here, and the ChatGPT will address your concerns and make a table accordingly.
I have always done this randomly since 1977. I was a kid but my mom and godmother were huge ERA supporters and it just seemed correct.
I tend to have my players interact with women or NBs because I’m self conscious about doing voices for men. But I’ve found that I definitely have some internal biases when it comes to certain things. Guard captains are almost always male, hospitality industry is usually female, etc etc.
Always trying to be aware of that and challenge those biases.
Another good question is about age, too. It seems like almost everyone is in the 20-40s range because that’s easier. But it’s a lot of fun to throw different age groups out there.
