In addition to the ethical considerations regarding appropriate ways to leverage and incorporate generative AI into your research and academic work, there are many other ethical concerns to be aware of. In particular, there are many ethical concerns regarding the data that is used to train generative AI models.
These include:
- Unknown data sources - many of the more popular generative AI models offer very limited transparency into the data that is used to train their models.
- Unauthorized inclusion of others works - despite the lack of transparency into the source and content of the datasets used to train generative AI models, many artists, authors, programmers, and others have discovered content generated by AI models that is nearly identical to their own. In many cases, there was never any authorization given to use their works as part of the training data.
- Rights & Attributions - in addition to including the works of others without authorization, some generative AI models imitate another creator without clearly stating so and offering proper credit and attributions. This can be of particular concern when using AI generated content in your research and academic work as you may be inadvertently including another person's work with incorrect or no citation and attribution.
- Bias - as with many AI models, generative AI models often include biases from the data they are trained on. Many of these biases in generative AI are yet to be fully identified and understood due to the newness of the technology and emerging capabilities. A few examples of biases that have been identified include:
- A tendency to use American/western perspectives in the responses.
- Amplifying societal stereotypes. This is particularly noticeable in image generators. Example: when prompted to create an image of a group of doctors the majority of people in the image are white males.