George FreemanThought Leader | Author | Architect| Florida, United States
A recent comment on this platform made me think about knowledge, perspectives of truth, and value. As a user stated, and I’m paraphrasing, “Authors on this platform occasionally share personal opinions, perspectives, and interpretations that are inconsistent with PMI standards, and content of this type should have no place in a knowledge base such as PMI Infinity.”
Although this statement concerns me, it raises a fair and important question that deserves challenging, not just in the context of project management and PMI Infinity but in general regarding generative AI tooling and service offerings, and that is:
- Does knowledge provided to a consumer only find value and purpose if an authority has determined its validity in advance?
If it does, then what is the purpose of: [1] Perspectives (i.e., something shared from one’s experiences) [2] Interpretations (i.e., the meaning given to experiences and events) [3] Opinions (i.e., what you think about something)?
Let me ask: [A] Would the learning experience provided by this community be possible without these three elements—I think not. [B] What is of greater value: The “good practices” provided by bodies of knowledge or the perspectives, interpretations, and opinions of practitioners who execute these practices in context and provide insight not necessarily covered in print—the answer is that the learning experience provided by both are of equal value.
So, my question to you is the following: * Should knowledge repositories (i.e., generative AI) contain only authoritative truth, or should they also contain truth that has its basis in perspectives, interpretations, and opinions?
Feel free to challenge the basis of these thoughts as well. Saving Changes...
Our understanding of the world and the decisions we make are shaped by the interwoven elements of knowledge, truth perspectives, and values. Our understanding of the world is based on the knowledge we have gained via education, life experience, and information. But truth isn't a fixed idea; rather, it depends on individual viewpoints shaped by life experiences, cultural contexts, and social conventions. Saving Changes...
When human beings are directly consuming content, I would definitely want to spread the net wide to ensure that different perspectives are being included.
The challenge comes when an AI tool which is expected to be relied upon to provide good predictions uses unvetted content as its source.
One way to solve the problem is to feed the AI tool all content, but have it apply some weighting based on the credibility of the source (e.g. PMI standards trump the feedback of a single user ; a large number of users trumps the feedback of a single user).
At present, tools like ChatGPT provide single responses to most prompts. If they instead provided contrasting responses with confidence weights and references/sources (the way that Copilot/Bing does), the risk would be reduced.
Kiron
...
1 reply by George Freeman
Feb 26, 2024 9:57 AM
George Freeman
...
Vetting is a necessary accountability practice but carries a heavy burden of bias when the process is non-transparent or not structured to be accountable in and of itself.
The battle (or challenge) largely exists in the minefield of policy narratives (not necessarily political), whether formed in public or corporate domains. These narratives, as appropriate as they may or may not be, will carry some form of managerial bias. Which, in most cases, is understood as a policy or directive, which is business as usual.
However, we may have an ethical concern when managerial bias impacts the publicly stated purpose and value proposition of content generated or distributed by an entity. NOTE: I’m saying this as a principled concern and not implying that this is occurring in the PMI domain.
All in all, transparency and accountability should and hopefully will guide the way forward.
Saving Changes...
William M Hayden JrAdjunct Assistant Professor| University at Buffalo, School of Management, Operations Management & StrategyBuffalo, Ny, United States
Thanks Kiron. Re: "PMI standards trump the feedback of a single user; a large number of users trumps the feedback of a single user)."
Q. Why would we immediately discount/void the input of a single user?
In the words of Yogi Bera: "When you come to a fork in the road, take it!"
Back around 1985 or so, I took a journey down that fork. While most all others were looking at tools of quality to improve project outcomes, I researched and learned that most projects fail due, not initially to tech, but to the fact engineers were never educated why, what, and how "to play nice with others." Cheers, Bill Saving Changes...
George FreemanThought Leader | Author | Architect| Florida, United States
Feb 26, 2024 7:12 AM
Replying to Kiron Bondale
...
George -
When human beings are directly consuming content, I would definitely want to spread the net wide to ensure that different perspectives are being included.
The challenge comes when an AI tool which is expected to be relied upon to provide good predictions uses unvetted content as its source.
One way to solve the problem is to feed the AI tool all content, but have it apply some weighting based on the credibility of the source (e.g. PMI standards trump the feedback of a single user ; a large number of users trumps the feedback of a single user).
At present, tools like ChatGPT provide single responses to most prompts. If they instead provided contrasting responses with confidence weights and references/sources (the way that Copilot/Bing does), the risk would be reduced.
Kiron
Vetting is a necessary accountability practice but carries a heavy burden of bias when the process is non-transparent or not structured to be accountable in and of itself.
The battle (or challenge) largely exists in the minefield of policy narratives (not necessarily political), whether formed in public or corporate domains. These narratives, as appropriate as they may or may not be, will carry some form of managerial bias. Which, in most cases, is understood as a policy or directive, which is business as usual.
However, we may have an ethical concern when managerial bias impacts the publicly stated purpose and value proposition of content generated or distributed by an entity. NOTE: I’m saying this as a principled concern and not implying that this is occurring in the PMI domain.
All in all, transparency and accountability should and hopefully will guide the way forward. Saving Changes...
Without some attempt to weigh one source of data against others, it is quite possible that a small number of highly vocal but fringe views could sway a prediction engine to providing skewed results. A good example is what we see with the "religious wars" surrounding adaptive ways of working One would hope that with a large enough data set this won't happen but unfortunately while there are hundreds of thousands of members in this community only a very small fraction are content contributors.
To use some statistics terminology, I would say that AI should focus on the middle of the distribution, but not ignore the tails.
When performing research without AI, literature studies are often used to identify what are the most widely held and best established beliefs on some subject. Understanding the contrarian views are still important because a) sometimes they're valid and just haven't caught on and b) it's often helpful to understand the positions of the people who don't agree with the general consensus.
A problem there is that the contrarian views often get disproportionate attention, something I learned in competitive debate. If the majority claims the earth is a ball, the person who claims it is flat gets to argue against everyone in the majority, amplifying that voice. That leads to Kiron's point that it may be necessary to weight the validity of the contrarian viewpoints that form the tails of the probability distribution. Saving Changes...
