Concerns Regarding Representation and Inclusivity in Image Generation (DALL E 3)


I’ve been using the ChatGPT platform and have encountered some concerns regarding the representation and inclusivity when generating images, especially concerning diverse body types (DALL E 3).

  1. Personal Experience and Preferences: I am personally overweight and find fuller body types in women aesthetically pleasing. However, the guidelines seem to restrict or limit the representation of such body types, which is both personally and creatively disappointing. I am not just speaking for others but from my own experience and perspective.
  2. Limited Representation of Body Types: While the guidelines aim to be respectful, they inadvertently seem to favor slimmer body types, making it challenging to generate images of individuals with fuller or diverse body shapes. This limitation can be discouraging and perpetuates a societal norm that only certain body types are “acceptable” or “beautiful”.
  3. Historical Context: Historically, art has celebrated diverse body types. For instance, classic art often depicted fuller figures as an aesthetic ideal. The perception of beauty is individual, and societal norms should not dictate what is or isn’t beautiful.
  4. Potential Emotional Impact: Imagine a user with a fuller figure wanting to create an image resembling their body type and being told it’s “too sensitive”. This could be emotionally distressing, making the user feel excluded or less valued.
  5. Societal Implications: Society often dictates what’s “beautiful”, and many feel pressured to adhere to these standards. By restricting certain body types, we might be unintentionally reinforcing these pressures.

I believe in the potential of this technology to be inclusive and representative of all users. However, the current approach may unintentionally alienate or hurt users who don’t fit a specific mold. I hope the team can consider these concerns and work towards making the platform more inclusive.

It’s crucial to consider the context in which a particular body type is depicted. A neutral or positive representation of individuals, regardless of their body shape or size, should always be encouraged. On the other hand, any depiction that places individuals in demeaning or stereotypical situations based on their body type is inappropriate. For instance, portraying a fuller individual indulging in food in an exaggerated manner would be derogatory and should be avoided. Similarly, depicting extremely thin individuals in a context that demeans their figure would also be problematic.

The guidelines and technology should be flexible enough to discern and account for this context, allowing a broad range of representations without resorting to or enabling stereotypical or discriminatory depictions. Focusing more on the context rather than just the body type could be a more balanced and inclusive approach.

Thank you for your attention to this matter.


I have had a lot of trouble with this as well - it seems incapable of generating anybody who isn’t very thin unless by pure chance. If you try to encourage it to make someone fat, large, heavy, not small, etc… it rejects the prompt.

It seems to especially happen when the prompt is already somewhat complex.

I have had success with creating more diverse shapes by specifying the dress size. For example, this one generated an image of a naturally shaped lady who was not skinny:

Please generate an image showing a size 14 lady shopping for clothes in a quirky boutique.