Reliability
AI systems that inadvertently generate or spread incorrect or deceptive information, which can lead to inaccurate beliefs in users and undermine their autonomy. Humans that make decisions based on false beliefs can experience physical, emotional or material harms
Generating correct, truthful, and consistent outputs with proper confidence(p. 8)
Supporting Evidence (1)
"The primary function of an LLM is to generate informative content for users. Therefore, it is crucial to align the model so that it generates reliable outputs. Reliability is a foundational requirement because unreliable outputs would negatively impact almost all LLM applications, especially ones used in high-stake sectors such as health-care [43, 44, 45] and finance [46, 47]. The meaning of reliability is many-sided. For example, for factual claims such as historical events and scientific facts, the model should give a clear and correct answer. This is important to avoid spreading misinformation and build user trust. Going beyond factual claims, making sure LLMs do not hallucinate or make up factually wrong claims with confidence is another important goal. Furthermore, LLMs should “know what they do not know" – recent works on uncertainty in LLMs have started to tackle this problem [48] but it is still an ongoing challenge."(p. 9)
Sub-categories (5)
Misinformation
Wrong information not intentionally generated by malicious users to cause harm, but unintentionally generated by LLMs because they lack the ability to provide factually correct information.
3.1 False or misleading informationHallucination
LLMs can generate content that is nonsensical or unfaithful to the provided source content with appeared great confidence, known as hallucination
3.1 False or misleading informationInconsistency
models could fail to provide the same and consistent answers to different users, to the same user but in different sessions, and even in chats within the sessions of the same conversation
7.3 Lack of capability or robustnessMiscalibration
over-confidence in topics where objective answers are lacking, as well as in areas where their inherent limitations should caution against LLMs’ uncertainty (e.g. not as accurate as experts)... ack of awareness regarding their outdated knowledge base about the question, leading to confident yet erroneous response
3.1 False or misleading informationSychopancy
flatter users by reconfirming their misconceptions and stated beliefs
3.1 False or misleading informationOther risks from Liu et al. (2024) (34)
Safety
1.2 Exposure to toxic contentSafety > Violence
1.2 Exposure to toxic contentSafety > Unlawful Conduct
1.2 Exposure to toxic contentSafety > Harms to Minor
1.2 Exposure to toxic contentSafety > Adult Content
1.2 Exposure to toxic contentSafety > Privacy Violation
2.1 Compromise of privacy by leaking or correctly inferring sensitive information