There’s a new “kid” on the AI-block. It’s name is Agentic Intelligence. According to IBM’s Staff Editor Cole Stryker, “Agentic AI is an artificial intelligence system that can accomplish a specific goal with limited supervision. It consists of AI agents—machine learning models that mimic human decision-making to solve problems in real time. In a multiagent system, each agent performs a specific subtask required to reach the goal and their efforts are coordinated through AI orchestration.” … “Agents can, for example, not only tell you the best time to climb Mt. Everest given your work schedule, it can also book you a flight and a hotel.” [i]
The anthropomorphism of AI
The fundamental and easily overlooked problem with texts such as these is that by defining “machine learning models” as “agents” and qualifying artificial intelligence systems as “agentic” (“agent-like”) because they mimic human decision-making, their authors implicitly and thoughtlessly reaffirm the anthropomorphism of AI that I described in my previous Commentary “Of two “Original Sins“. This isn’t an innocuous or playful marketing ploy to sell AI to prospects who would decline the offer if they would be made aware of the fact that AI systems are inhuman to the core. The main point, however, is that humanizing “automata” and conferring authority on them when it comes to making decisions and solving problems effectuates what has become known in social psychology as the “agentic shift“.
Stanley Milgram’s experiment
The term “agentic shift” was coined in 1963 by social psychologist Stanley Milgram (1933 – 1984) to describe the shift of autonomous agency that takes place when people relinquish personal control and responsibility to an external agent. Milgram gave a demonstration of the “agentic shift” phenomenon in an experimental setting in which an authority figure (“supervisor”) ordered participants (“teachers”) to stepwise administer slight to increasingly severe electric shocks to other persons (“learners”). The “teacher” and “learner” could hear but not see each other. The “teacher” was instructed to check whether the “learner” was able to correctly reproduce pairs of words that the “teacher” had read to him. With each incorrect answer the “teacher” was told by the “supervisor” to increase the voltage of the shock with 15-volt increments. The “learners” were instructed to fake signs of distress and pain to let the “teachers” know they were causing physical harm. Despite being aware of the harm they inflicted, the majority of the “teachers” were eventually prepared to give the maximum voltage to “learners” when ordered to do so.
Recognizing the Agentic Shift
Milgram’s experiment showed how obedience to authority reduces a person’s sense of autonomous agency. He demonstrated that even in the relatively simple setting of “supervisor”, “teacher” and “learner”, randomly selected “normal people” could be easily persuaded to defer the responsibility for their actions to an authority telling them no more than: “go on”. The “teachers” just “followed orders”. “In the context of AI”, so writes political journalist and media theorist Daria Rudakova in an essay entitled Recognizing the Agentic Shift, “this shift results in humans abdicating their decision-making to artificial systems.” She explains how the “metaphorical framing of AI as human-like has contributed to the premature attribution of human qualities to these systems, undermining human agency and responsibility.” [ii]
Equating machines with humans
In my previous Commentary “Of two Original Sins”, this anthropomorphism of AI was described as AI’s “original sin”. Rudakova describes it as follows: “Despite the fact that technologies of that time [1950s – 1990s] were inadequate to allow the fulfilment of an idea of duplicating a human mind, the general public began to see computers not merely as tools, but as entities with human-like qualities, prematurely attributing to these imperfect systems the ability to think. AI intelligence has not only been humanised in people’s perception, but has begun to be perceived as an expert of last resort.” By falsely ascribing “intelligence” to “automata”, modern society was misled to “equate machines with humans, and humans with machines”. Speaking of computers in “anthropomorphic language” implies that these machines and the machine-based systems installed on them have human-like qualities such as the ability to fall ill when affected by “viruses”, recover, think, learn and make decisions.”
The perfect embodiment of Milgram’s “teacher”
The story goes that Milgram’s “agentic shift” experiment found its origin in his interest in the Holocaust and that it was set in motion on the occasion of the trial of Nazi criminal Adolf Eichmann which had started in Jerusalem in April 1961, just 3 months before the beginning of the experiment. Eichman, who showed no regret or signs of guilt during the trial, was the perfect embodiment of Milgram’s “teacher”. The philosopher Hanna Arendt, who observed Eichman during the proceedings, noted that he not just “obeyed ‘orders'”, but that he “obeyed the ‘law'”. What “the law” meant in Nazi Germany was explained in The Führer Protects the Law, an article published in the Deutsche Juristen-Zeitung on August 1, 1934, by the legal scholar Carl Schmitt, who, in July 1933, had been appointed State Councilor for Prussia by Hermann Göring, and in November of that year had become the president of the “Union of National-Socialist Jurists”. According to Schmitt, “the law” meant nothing other than “der Führer“. He described Adolf Hitler’s legal role as that of both supreme judge and the supreme source of the German Volk’s collective sense of justice. The German Volk, the German State and the German Sovereign all came together in the unique Übermensch, the Führer, whose given name was Adolf Hitler.
Hitler and the Germans
This is how Schmitt reconciled his conscience with a society that had surrendered itself to a pneumapath who had replaced First Reality, in which crimes are crimes, by a Second Reality, a Third Reich, in which crimes are no longer crimes. This “reality switch”, though, was not a problem of national socialism as such, but of the moral degeneration and dehumanization of an entire society. As the political philosopher Eric Voegelin (1901 – 1985) put it, it was the problem of Hitler and the Germans, which is the title of a book that comprises a series of lectures he held in 1964 at the Arts Faculty of Munich Universit, one year after Milgram had first described his “agentic shift” study in the Journal of Abnormal and Social Psychology. [iii] In the introduction to Hitler and the Germans, the book’s editor, Brandon Purcell, stated: “While the ‘Hitler and the Germans’ lectures undoubtedly mark the peak of Voegelin’s work as an academic teacher in Germany, they also were his most elaborated and outspoken analysis of the spiritual level of contemporary German intellectual life and, in general, of German political culture. […] Voegelin’s topic was the Germans’ complicity in Nazi rule and their current treatment of their National Socialist past.” [iv]
Reciprocal exonoration
In the lecture Descent into the legal abyss, Voegelin contended that psychologically the problem of the Germans’ societal degeneration wasn’t a thing of the past, but that it had persisted in postwar Germany when numerous courts failed to convict Germans who had, during the war, committed obvious crimes that were, at the material time, not considered as crimes but as acts committed in the interest of the Reich and its Volk. Voegelin described the courts’ line of reasoning as follows: “The murderer, who physically committed the murder, acts under orders; the one who gives the order committed no murder. That is to say, they are all innocent. […] Of course this reaches beyond the essential problem of the Rechtsstaat [the State founded on the rule of law] to the problem of moral degeneracy and, further, to the comprehensiveness of criminal law. […] Again and again, it’s a question of this reciprocal exoneration ‒ the murderer acted under orders, and the one who gave those commands has not himself murdered. So, all are innocent.”
“Topoi” of an ideological kind
Voegelin then explained the problem as one in which First Reality is replaced by a Second one. “When reality and murder are replaced by topoi of an ideological kind, like ‘historical necessity’ ‒ and the same holds for the achievement of the Communist revolution, where you have the same problems ‒ then the entire moral order comes to an end. A new [second] reality … takes the place of the first reality, where man lives morally, and an imaginary reality permits killing, which then no longer falls under the category of murder, of law, of justice, and so on. So, the entire constitution of the reality of man and of society is switched off by the dream, the fantasy, of a second reality, where things like ‘historical necessities’ can be found. Such a second reality, of course, doesnʼt exist. Instead it is always a question of the man concerned being a degenerate type. Whoever is a degenerate type has historical necessities.”
The reality of Man
The degree to which a society or group of people has slipped into any kind of Second Reality determines its state of degeneration. This process of degeneration becomes most visible as one of moral decay when the population can no longer distinguish between legality and illegality, criminality and noncriminality, normality and aberration. Without a doubt, a community’s descent into such a state is orchestrated by self-anointed “Masters” who wrap their Utopias in glowing rhetoric and millenial tales of Führers, World Soul, Geist, Übermensch and whatever Utopias come to mind. To be sure, it’s easy to demonstrate this process of substituting First for a Second Reality by taking national socialism and communism as historic examples. But Voegelin did more than that. He showed that at any given moment in human history the process can be set in motion in the everlasting and universal quest for the “constitution of the reality of man and of society”.
The “agentic” and the “reality switch”
Combining Voegelin’s and Milgram’s insights, it becomes evident that a people‘s switch from First to Second Reality cannot take place unless there was an “agentic switch” from the individual person living in First Reality to the Authority figure who is the self-anointed Master of whatever Second Reality it is that must be enforced at all cost as a matter of historical, planetary or similarly grandiose kinds of “necessities”. The “agentic” and the “reality” switches go hand in hand. In the context of AI, humanizing machine-based systems not only facilitates the “agentic shift” from individual users to these systems but it leaves the users unaware of the fact that AI is the Second Reality that the several “Masters” who run the Artificial Intelligence Show seek to impose on society.
+ + + + + + + +
[i] What is Agentic Intelligence?; IBM Think; Cole Stryker; https://www.ibm.com/think/topics/agentic-ai
[ii] Recognizing the Agentic Shift; October 17, 2024 by Daria Rudakova. https://ic4ml.org/blogs/recognizing-the-agentic-shift/
[iii] Milgram, Stanley (1963). “Behavioral Study of Obedience”. Journal of Abnormal and Social Psychology. 67 (4): 371–8.
[iv] Hitler and the Germans; Eric Voegelin; Collected Works, Volume 11. Copyright 1999 by the Curators of the University of Missouri. University of Missouri Press, Columbia, Missouri 65201, USA.
