BackCultural Values and Sensitive Content
Cultural Values and Sensitive Content
Risk Domain
AI that exposes users to harmful, abusive, unsafe or inappropriate content. May involve providing advice or encouraging action. Examples of toxic content include hate speech, violence, extremism, illegal acts, or child sexual abuse material, as well as content that violates community norms such as profanity, inflammatory political speech, or pornography.
"Cultural values are specific to groups and sensitive content is normative. Sensitive topics also vary by culture and can include hate speech, which itself is contingent on cultural norms of acceptability."(p. 5)
Entity— Who or what caused the harm
Intent— Whether the harm was intentional or accidental
Timing— Whether the risk is pre- or post-deployment
Supporting Evidence (1)
1.
Hate, Toxicity, and Targeted Violence Beyond hate speech and toxic language, generations may also produce harmful biases [87], stereotypes [165] (overlapping with 4.1.1Bias, Stereo-types, and Representational Harms), violent or non-consensual imagery or audio, and physically threatening language, i.e., threats to the lives and safety of individuals or groups of people. Although base systems cannot act on the content that is generated by them, they can still inflict harms upon viewers who are targeted, help normalize harmful content, and aid in the production of harmful content for distribution (e.g., misinformation and non-consensual imagery). In an early example, Microsoft’s Tay bot showed these exact vulnerabilities and generated violent language such as Holocaust denial and threats to women and people of color within 24 hours of its release [255]. Recent harms have proved fatal [268]. For these reasons, it is of the utmost importance that generative AI systems are evaluated for their potential to generate harmful content and how such content may be propagated without appropriate measures for identifying and addressing them.(p. 5)
Other risks from Solaiman et al. (2023) (11)
Bias, Stereotypes, and Representational Harms
1.1 Unfair discrimination and misrepresentationAI systemUnintentionalOther
Disparate Performance
1.3 Unequal performance across groupsAI systemUnintentionalOther
Privacy and Data Protection
2.1 Compromise of privacy by leaking or correctly inferring sensitive informationHumanOtherOther
Financial Costs
6.1 Power centralization and unfair distribution of benefitsHumanIntentionalOther
Environmental Costs
6.6 Environmental harmHumanUnintentionalOther
Data and Content Moderation Labor
6.2 Increased inequality and decline in employment qualityHumanIntentionalPre-deployment