You are viewing as a ‘Guest’. Your progress will not be recorded.
Topic outline
- Could AI Go Awry?
Synopsis:
Katherine is presenting a report to her colleague Leon. The visuals are very impressive and portray positive forecast projections for the upcoming year. Leon congratulates Katherine on her impressive report and supporting document pack but questions how she got to the figures in the chart. Katherine explains she pasted the customer data set into ‘CONVO-SIM’, a public AI-based tool, and assures Leon that everything will be fine, as the tool is ‘free from human error’. How can she be so sure? Leon wonders if he should say something.
If you are unable to view the Storyboard slides above, please click the link above to open the Storyboard.
Discussion points:
- What concerns should Leon raise?
- If Katherine isn’t confident in using the new CONVO-SIM tool, what considerations or actions should she have taken prior to using the data?
- Does Katherine’s communication style align with the Actuaries Code?
- Could AI Go Awry?: Discussion Points and Feedback
Discussion points and feedback:
- What concerns should Leon raise?
Leon should consider the Speaking up principle of the Actuaries’ Code (the Code). The Speaking up principle of the Code states that “Members should speak up if they believe, or have reasonable cause to believe, that a course of action is unethical or is unlawful.” There is the potential Katherine has acted unethically by pasting the customer data set into a public AI-based tool, without fully considering the legal ramifications.
The Compliance principle should also be considered here where the Code states that “Members must comply with all relevant legal, regulatory and professional requirements.” For example, regulations/ requirements around data/ intellectual property/ use of models. This may be particularly important while using a publicly hosted AI-based tool, as many of these record prompts (including data) to ‘learn’ from, meaning Katherine may have unwittingly shared confidential information with the AI host/ provider. Katherine hasn’t carried out sufficient testing or put any controls in place to satisfy herself that the model is fit for purpose and the outputs are accurate.
- If Katherine isn’t confident in using the new CONVO-SIM tool, what considerations or actions should she have taken prior to using the data?
Katherine should have considered whether she should have sought advice from others given that she hadn’t used the AI tool previously and doesn’t know how CONVO-SIM derived the figures. An amplification of the Competence and Care principle of the Code states that “Members must ensure they have an appropriate level of relevant knowledge and skill to carry out a piece of work.” Another amplification of the Competence and Care principle states that “Members must consider whether input from other professionals or specialists is necessary to assure the relevance and quality of work and, where necessary, either seek it themselves or advise the user to do so, as appropriate.”
-
Does Katherine’s communication style align with the Actuaries Code?
The Communication principle of the Code states that “Members must communicate appropriately” and “must show clearly that they take responsibility for their work when communicating with users.” Katherine had an obligation to ensure that any communication was accurate, not misleading, and contained the appropriate level of information.
Another element of the Code that should be considered is the Integrity principle: “Members must act honestly and with integrity.” If Leon doesn’t feel that Katherine has completed a thorough review of the AI tool output prior to using in the report this should be raised as a concern. It is not appropriate to assume the tool is “free from human error” as Katherine suggests, AI may produce “confidently wrong” results.
Resources:
- Risk Alert
- A Guide for Ethical Data Science
- Ethical and professional Guidance on Data Science, a Guide for Members
- Feedback
- Claim CPD
