1.2.1. Bistability: Two-Sided Stories
First, the bistable information from the construction of a context shared by two or more agents universally generates two-sided stories (e.g., ”he said, she said”). Or, different stories arise from observers and actors [9
]. Or, bistable images (Figure 1
) and interpretations exist. World views can persist (e.g., religious or atheist, conservative or liberal, myths, or fantasies) when bistable views are sufficiently stable, promoted, or controlled. If sufficiently stable, opposing views can evolve into different tribal customs with supporting functions and hierarchies [10
]. Bistable views promote competition among views that forces them to evolve. e.g., “Fracking Buzzwords Evolve, From ‘Ramp Up’ to ‘Capital Discipline’ ” (in [11
]). However, few humans like to exchange two-sided or contradictory views, motivating the replacement of disliked views with conformity to a favored view; e.g., humans have made extraordinary advances with quantum theory, yet a consensus regarding its meaning remains elusive a century after its discovery, which is underscored by attempts to supplant it with a one-sided rational theory [12
] identifies several examples of rational choice theory in cooperative situations throughout nature, but he states that the theory breaks down under conflict, as with competition where interdependence theory flourishes. Darwin [10
] found that human groups cooperated to offset competition in their “struggle for existence”. Competition leads to the best ideas [14
]; it supports Justice Ginsburg’s ([15
], p. 3) argument for an “informed assessment of competing interests”. Competition exposes weaknesses, motivating mergers. (e.g., United Technologies and Raytheon announced a merger to improve their ability to innovate for national security [16
]). Similarly, competition among bacteria can lead to an “awareness” of a genetic deficiency that forces horizontal gene transfers (mergers) that offer “protection from predators, pathogens, and environmental stress” (in [17
: Competition drives innovation [18
], but it flounders under the minority control of command decision-making regimes. Consider that China’s R&D expenditures are second in the world to the United States (U.S.) [19
]. However, China’s command directed finance (single-sided views), its weak intellectual property protection, and its rampant corruption impede innovation (e.g., [20
]). As General M. Hayden, the former Central Intelligence Agency (CIA) and National Security Administration (NSA) chief, stated to his Chinese counterparts on the innovativeness that has so far eluded China: “You can’t get your game to the next level by just stealing our stuff. You’re going to have to innovate” [21
]. To better understand the barriers to innovation in China, Taplin [22
“Small private-sector firms often only have access to capital through expensive shadow banking channels, and risk that some better connected, state backed firm will make off with their designs— with little recourse.”
For similar reasons, making movies in China has begun to flounder [23
"Tougher government censorship has blocked potential hits and compelled filmmakers to stick with safe formulas that aren’t winning audiences... In the year after Chinese President Xi Jinping put the Communist Party’s propaganda office in charge of regulating films, China’s box-office totals are headed for their first annual decline in at least a decade."
1.2.2. The Measurement Problem
The second effect from interdependence is a measurement problem that we associate with the duality of behavior and imagination, orthogonality, and multi-tasking. This problem commonly negates the value of self-reported questionnaires (e.g., the irrelevance of self-esteem and academics; in [24
]; ego-depletion, likely the most famous theory to fail in social psychology; in [25
]; and, the mis-informed views of managers about their own organization’s performance; in [26
]). The measurement problem is aggravated regarding censoring one world view in favor of another’s, causing individuals, teams, organizations, or the tribes that act on single-minded decisions to be more likely associated with error (e.g., Boeing’s 737 Max accidents; in NTSB [27
]; also, in the case study, we review the remediation of the environmental destruction caused by the U.S. Department of Energy’s (DOE’s) mismanagement of radioactive wastes).
The measurement problem reveals social research to be largely inapplicable to the design of autonomous systems (HMTs). The participants in teams can be counted, but measuring weight and mass have little impact on team autonomy; and, the entropy of a team’s configurations can be measured as well as the flow of information and energy that produces teamwork, but is that sufficient? Using the transformation of water to ice as an analogy initially showed promise for a system’s reduction of configurational entropy generalized to HMTs (agreeing with [5
]), but it was insufficient, leading us from hurricanes to bio-crystals and living engines to better capture the thermodynamics of teams. However, these methods ignore a chief characteristic of humans: imagination [28
. Questionnaires capture beliefs and imaginations well, but not individual behaviors [29
]. We can measure the decisions that humans make, but human justifications are often unrelated to decisions (Tversky, in [30
]). Likely related, many of the most important findings from social science are based on surveys, but the results have proved to be difficult to replicate [31
], which is a problem not exclusive to social science (e.g., in medicine, see [32
]; in statistics, see [33
]; in physical science, see [34
]). We explored the tension between beliefs and behaviors to draw two conclusions: surveys of cognitive beliefs are commonly unrelated to the behaviors that they purport to predetermine (e.g., implicit racism; in [35
]); and this failure probably occurs because observation and action are orthogonal [18
]. The former impedes the generalization of social theory to HMTs; but, surprisingly, the latter might be good news for social theory.
], the founder of group dynamics, concluded that interdependence caused a whole (e.g., a group) to be greater than the sum of its parts, which suggests a loss of information from the simple sum of contributions by individual team members (see also [37
]). However, one of the earliest theories of interdependence, that a complementarity of opposites (orthogonals) formed the best human relationships, failed to garner statistical support [38
]. We conclude that interdependence among humans, the phenomenon that transmits social effects, misleads intuition at human levels that are similar to entanglement at quantum levels, causing a loss of information from measurement. Notice a similarity between interdependence and the quantum [39
Our ordinary intuition regarding physical systems is that, if we know everything about a system, that is, everything that can in principle be known, then we know everything about its parts. However, Einstein explained to Bohr—in quantum mechanics, one can know everything about a system and nothing about its individual parts.
Correlations present three problems to interdependence theory: First, orthogonal implies independence, but a zero correlation also implies independence. Second, however, for a close relationship, a correlation means a state of interdependence, mutual relationship, or connection between two or more social elements, for which the evidence does not support (e.g., [40
]), a conundrum. We often hear that a correlation does not mean causality; but how can we account for causality with interdependence without a correlation? Third, the absence of a correlation in close relationships has been taken to mean the absence of an interdependent relation between orthogonal social objects. However, if two social objects are interdependent, they are causally connected. Interdependence orders two or more social variables. If a pair is in an interdependent relationship that occupies orthogonal roles, the information they collect and self-report, coming from orthogonal roles, has zero correlation by definition.
], p. 32) claims that humans live in a dual reality, one objective and one imagined. This dual reality includes the roles that humans play. From Shakespeare [41
], “All the worlds a stage, And all the men and women merely players … And one man in his time plays many parts.”
To play these many parts requires different interpretations of reality; e.g., a waitress-cook pair working together is interdependently connected, yet each self-reports that each sees the world differently. Another example of orthogonality and the lack of a correlation is found for the carpus collosum, the two independent halves of a brain work seamlessly together like a team, but once surgically severed into two parts, each half of the brain sees and self-reports the world differently [42
]. Instead of independent elements, the same statistical effect occurs if the roles of a team are orthogonal (ship’s captain, ship’s engineer, etc.).
Let us accept that social theory is predicated on the individual; but, teams are comprised of interdependent individuals [4
]; and, that a theory of complementarity for interdependent human couples has failed due to of negligible correlations, indicating no relation or independence [18
]. However, if we accept that a state of interdependence is proof of a strong correlation, how to resolve this paradox? If team members are interdependent, they are causally connected; e.g., husband-wife; prosecutor-defense attorney; and, pitcher-catcher-outfielder. The conundrum resolves if information is generated in orthogonal roles, if information is collected from individuals, and if only one interpretation of reality can be processed at a time (i.e., individuals are poor at multitasking; in [43
]), then interdependence organizes a team of agents into orthogonal roles, but is observed as individuals.
Individuals multi-task poorly [43
]; e.g., using a cell-phone while driving a car. In contrast, multitasking is the function of teams [8
], making teamwork an emergent property that is unequal to the sum of a team’s individual contributors. Forming and operating a team requires multiple streams of communications (verbal, non-verbal) that include the constructive and destructive interference transmitted by interdependence; e.g., destructive interference among humans includes angry debate, even among the best teams [45
]; constructive interference includes support for the behaviors that are chosen for a context, for which AI systems, especially machine learning, have been unable to manage so far [46
], a problem that must be solved for effective and efficient human-machine teams (see our Special Issue in AI Magazine; [1
We can model orthogonality with the dot product of two eigenfunctions A
where for the Dirac delta,
if the two eigenfunctions, A
, are identical, 0 otherwise. (Using word2vector, the words democracy, freedom, and majority associate; consensus, unity, and sociology associate; but, these two groups do not.)
1.2.3. Non-Factorable Information
Third, the final effect of interdependence is non-factorability (e.g., the intractability of assigning blame in complex legal cases). Humans work with non-factorable information in courtrooms, in investigations, politics, making movies, tradeoffs, etc. The adversarial approach in the courtroom is the method that humans traditionally use to create a context where the truth can be sought or hidden information exposed like biases (e.g., confirmation bias, in [47
]). As an example of a trade-off, from signal detection theory, Cohen [48
] reported that a "narrow waveform yields a wide spectrum, and a wide waveform yields a narrow spectrum and that the time waveform and frequency spectrum cannot both be made arbitrarily small simultaneously."
As a tradeoff, we can know the effect of interdependence, as it acts on the whole (say a team), or its parts (the members of a team), but not both simultaneously [18
]. That is why constructing a theory of interdependence proved to be “bewildering” ([49
], p. 33).
] first mathematically predicted that a tribe adopts a view of reality that is different from a neighboring tribe. What is the evolutionary advantage of having two tribes with different interpretations of reality? From Simon [50
], if every story represents a bounded rationality, but if at least two sides exist to every story, then it takes at least two sides to determine social reality, which sets the stage for adversarial or competitive determinations of reality [8
]. When neutral agents do not exist in an adversarial context, conflict becomes likely [51
]. However, in the U.S., the largest political group is the collection of voters known as independents or neutrals [52
]. Neutrals serve two functions: first, when political parties attempt to persuade neutrals to join their side, political conflict is moderated; and, second, like a quasi-Nash equilibrium (each side holds opposing, stable, inflexible beliefs), two political parties drive neutrals to process the discordant information that was generated by the opposing parties, while the neutrals justify their votes. An outcome is a political side that wins, a jury’s decision, or the formation of a scientific consensus. Another possible outcome is a compromise. Neither side likes a compromise, but a compromise can last if judged to be fair [53
As a sidebar, Nash’s [54
] equilibrium solves the non-cooperative Prisoners Dilemma Game, which is purposively structured to provide a higher payoff between cooperators than competitors, justifying Axelrod’s [55
] claim that cooperation is superior to competition. Instead, we elect to see the Nash equilibrium as the only method for resolving non-factorable information (later, we rename this phenomenon as a social harmonic oscillator). Witness that Nash equilibria do not occur in public venues under authoritarian regimes; in contrast, the adversarial competition that is expressed by a Nash equilibrium is the source of justice in the courtroom [56
], the best test of a political question [15
], the means to the best ideas in speech [14
], and the primary route towards innovation [18
] proposed an adversarial approach for the US Navy as a way to reduce accidents at sea by subjecting decisions under uncertainty to a vote. Recently, the adversarial approach solved the Uber self-driving accident that killed a pedestrian in March 2018 [1
]. At first, the Police Chief blamed the pedestrian, who was supposedly on drugs at the time. Next, the software that was used by Uber was blamed because the company struggled to average 13 mi before a human observer had to intervene (by way of a comparison, the Waymo self-driving car averaged about 5,600 mi per human intervention). Finally, NTSB [3
] blamed the company for disconnecting the car’s emergency breaks, which had correctly sent a signal to apply braking to its disconnected brakes.
However, how to model the adversarial approach if we cannot explain causality with AI [53
]? Part of the problem is the non-factorable effects of interdependence on the social interaction [8
] making the rational determination of social causality impossible (e.g., the Syrian war; a nasty divorce; and, hostile mergers) (an example of a hostile merger is Barrick’s bid for Newmont’s gold mining operations [58
]). As compared to AI machines that are excellent at classification, even if they cannot explain their decisions, (e.g., discovering incorrect prescriptions; in [59
]), our goal for this work-in-progress is to seek a foundation for AI to be able to manage aspects of causality to achieve the best social interpretation or social decision by an A-HMT. (e.g., with opposing case-based reasonings; in [60
Before ending this introduction to interdependence theory, it is instructive to the reader that Social Psychologists seek to remove the effects of interdependence in experimental data to better replicate their findings ([61
], p. 235, by obtaining independent and identically distributed data), as also recommended by engineers [62
] and information theorists [63
]. We have countered that removing interdependence to amplify the value of individual choices has not only impeded a theory of teams, but it is also perhaps why social science research cannot be replicated [31
]. As an analogy, the removal of interdependence from the study of teams is akin to removing the “pesky” quantum effects to better study the atom.