There’s a wonderful article in the FT today about whether AI should have a conscience. It concludes
The shift to agentic AI — bots that don’t just assist but actually execute tasks and exert judgment — will raise the stakes. As AI gets more humanlike, and its role within the company gets more senior, how it responds to complex challenges and conflicts will matter more.
When is it better to ignore a command?
When could pursuing a short-term goal lead to longer-term problems?
When is it OK to tell the boss to “shove it”?
For better or worse, it’s really no different to what companies seek in their employees.
For less critical and more process-driven jobs, employers seek workers who follow rules. At senior levels, where an individual’s actions can affect the value of the whole firm, good judgment in unusual situations becomes valuable, and commands higher pay.
This reminds me of a conversation I had with my actuary friend Chris last night , while walking over the Thames to Waterloo. We had been discussing with another actuary friend Derek our wish to automate CDC’s pricing to a point that no judgement was required and technology managed fairness between all. You can read Chris’ thoughts in the Prof Pens/blog article here
What we’d concluded in an hour of rigorous debate is that you cannot automate fairness as well as you can use the judgement of smart people. This is why you need to have an actuary challenging the decisions of the other actuary and why you cannot just leave it to technology to find fairness.
I won’t explain the final detail as we have put the three of us under NDA. But I can say that a very large part of our conversation was about doing what was and is right in the eyes of an unseen but felt force. It does not surprise me that Chris, Derek and I found ourselves talking about religion.
Agentic AI was a step too far for the two actuaries I was with. Did I sense they were considering AI devaluing their human judgement? I suspect that they are like any other senior decision makers , believing they still know best. And so long as we have churches, we will have places to celebrate the human conscience and the superhuman force that governs it.
Conventional retirement planning answers the question: how much do I need to save to stop working at a given age? The model required must answer a different question: how do I guarantee a fully funded, high-quality final chapter of life — however long I live — while extending the productive years of middle life rather than compressing them?
Change the thinking that caused the present failure of pensions to serve many individuals.
Quis custodiet ipsos custodes?
I’m sure Chris and Derek are not guilty of this, but I have seen and heard a lot of groupthink practised by many actuaries.
I believe we need to have professional (preferably non-actuarial) and lay trustees willing to challenge the actuaries in the room on occasion.
At times I think any application of intelligence in actuarial standards and expectations would be an improvement. From my “22 years experience of Shared Ambition – A first of a kind??” presentation to Pensions UK East Midlands Group on Thursday I noted the following:
Inappropriate or misleading valuations obscured the amount of surplus until 2024
– Scheme actuary’s request to use asset based valuation model and single discount rate was rebuffed by TPR in 2008
– Not pursued for 2009 valuation as could not justify use of 12 month trailing dividend yield as a predictor immediately after the Global Financial Crisis
2021 Valuation (112% TP Funding) was particularly problematic in obscuring surplus
– Best Estimate assumed a scheme wide future investment return of c. 4.2% (CPI inflation assumption 3.6% p.a.)
– Solvency Valuation carried a 69% risk premium (Insurer’s profit margin ?) over best estimate
Trustee Assessed the 2024 Valuation Results against lessons learnt
– Concluded that whole scheme implied Technical Provision discount rate (±5.9%) was a prudent basis for expected investment performance and for surplus decision making
– Until re-assessed after the 2027 valuation this will be used for all decisions on surplus allocation and discretionary revaluations ignoring any subsequent volatility and as a base for investment decisions