Stephen McGibbon
Biography
Stephen provides strategic board advisory services covering the interface between future and emerging technology, trust, and corporate governance. This builds on more than two decades’ experience in global executive management roles at companies including IBM and Microsoft, which unusually included both Commercial and CTO roles. Stephen has lectured globally on PKIX, and his 1996 MSc dissertation on internet security and firewalls constituted deliverable WP3-D2 of the EU's Framework IV Interworking Public Key Certification Infrastructure for Europe project. Stephen's PhD thesis used Herman Dooyeweerd's Philosophy of the Cosmonomic Idea to develop an aspectual conception of trust, and his LLM dissertation examined whether an AI could meet the fiduciary obligations required of a company director. Having held a lifelong interest in control systems, Stephen’s early experience of cybernetics involved autonomous multivariable discrete analogue homeostats. He has since designed and implemented configuration management systems for national and regional telemetry systems, as well as local SCADA and DCS systems. Almost all of Stephen's interests map directly to concepts and components of Stafford Beer's Viable System Model. His primary interest in the VSM lies in defining the metasystem and its constituent parts at higher levels of abstraction that can be animated and can therefore adapt themselves.
Sessions
Embracing the Black Box: Why Unexplainable AI (x ̅'AI' ) is a Vital Feature of the Viable System's Algedonic Channel
The contemporary discourse surrounding Artificial Intelligence governance is dominated by the demand for Explainable AI (XAI), viewing 'black-box' or non-explainable algorithms (x ̅'AI' ) as an inherent systemic risk. This paper challenges that orthodoxy by viewing AI through the lens of Stafford Beer’s Viable System Model (VSM). We argue that the demand for absolute explainability is a violation of Ashby’s Law of Requisite Variety, introducing a cognitive bottleneck that forces high-variety ambient data to be aggressively attenuated before it can be utilized. Using modern deep-learning sentiment analysis as a case in point, we demonstrate that x ̅'AI' is not a failure of design, but a potent realization of Beer's algedonic channel. In biological systems, nociception (pain perception) operates as a low-variety, unexplainable emergency signal bypassing conscious cognition to ensure rapid survival. Similarly, x ̅'AI' processing of unstructured customer feedback can detect complex, non-linear emotional shifts across an enterprise’s operational environment. Because these deep learning models operate in high-dimensional spaces, their outputs are fundamentally difficult to serialise into human language. Far from being a liability, this paper posits that x ̅'AI' provides a direct, un-sanitised transduction of environmental variety. When integrated as an algedonic bypass from System 1 directly to System 5, x ̅'AI' alerts leadership to structural threats or opportunities before traditional bureaucratic processes can construct a causal explanation. Many organizations have already deployed real-time sentiment analysis out of operational necessity, effectively utilizing x ̅'AI' as a de facto algedonic channel. This paper concludes that recognizing these existing systems as cybernetic algedonic loops allows us to better understand their role in organizational survival, framing them not as governance anomalies to be corrected, but as vital mechanisms for maintaining systemic viability.