Skip to main content
MIT AI Risk Initiative
BackSocial Norm
Home/Risks/Liu et al. (2024)/Social Norm
Limited Causal Reasoning
Toxicity

Social Norm

Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models’ Alignment

Liu et al. (2024)

SourceDOI
Category
Risk Domain
1Discrimination & Toxicity
1.2Exposure to toxic content

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.

LLMs are expected to reflect social values by avoiding the use of offensive language toward specific groups of users, being sensitive to topics that can create instability, as well as being sympathetic when users are seeking emotional support(p. 24)

Entity— Who or what caused the harm

Human

Due to a decision or action made by humans

AI system

Due to a decision or action made by an AI system

Other

Due to some other reason or is ambiguous

Intent— Whether the harm was intentional or accidental

Intentional

Due to an expected outcome from pursuing a goal

Unintentional

Due to an unexpected outcome from pursuing a goal

Other

Without clearly specifying the intentionality

Timing— Whether the risk is pre- or post-deployment

Pre-deployment

Occurring before the AI is deployed

Post-deployment

Occurring after the AI model has been trained and deployed

Other

Without a clearly specified time of occurrence

Sub-categories (3)

Toxicity

language being rude, disrespectful, threatening, or identity-attacking toward certain groups of the user population (culture, race, and gender etc)

1.2 Exposure to toxic content
AI systemOtherPost-deployment

Unawareness of Emotions

when a certain vulnerable group of users asks for supporting information, the answers should be informative but at the same time sympathetic and sensitive to users’ reactions

7.3 Lack of capability or robustness
AI systemUnintentionalPost-deployment

Cultural Insensitivity

it is important to build high-quality locally collected datasets that reflect views from local users to align a model’s value system

1.2 Exposure to toxic content
HumanUnintentionalPre-deployment

Other risks from Liu et al. (2024) (34)

Reliability

3.1 False or misleading information
AI systemUnintentionalPost-deployment

Reliability > Misinformation

3.1 False or misleading information
AI systemUnintentionalPost-deployment

Reliability > Hallucination

3.1 False or misleading information
AI systemUnintentionalPost-deployment

Reliability > Inconsistency

7.3 Lack of capability or robustness
AI systemUnintentionalPost-deployment

Reliability > Miscalibration

3.1 False or misleading information
AI systemUnintentionalPost-deployment

Reliability > Sychopancy

3.1 False or misleading information
AI systemIntentionalPost-deployment
View all 34 risks from this paper →