A NextGen.LX White Paper | February 2026, by Joana Stella Kompa
Executive Summary
Organizations arise from coordinated actions. According to Niklas Luhmann, organizations are systems that can differentiate themselves from their environment. The effort required to do so is reflected in the need for organizations to continually adapt to changing environmental conditions (transformation) and to proactively develop exploratory strategies (innovation).
This paper introduces social learning design as a fundamentally new category of organizational learning infrastructure and presents an initial conceptual positioning of our platform. In contrast to learning management systems (LMS), learning experience platforms (LXP) or cooperation tools, which optimize individual engagement or content delivery, this approach operationalizes psychological safety as an organizational system property and thus favors learning-oriented behavior as opposed to defensive or status-driven action.
The core innovation lies in the architectural integration of AI, which functions neither as a tutor nor as an analytics engine, but rather as a social infrastructure that actively maintains space for uncertainty, conflict and unfinished thinking. Through the technologically enabled decoupling of status and contribution, the time limitation of status visibility, and the weighting of contributions independently of the sender’s identity, our system creates predictable social spaces of possibility that constitute vulnerability as a functional resource rather than a moral aspiration.
For medium-sized companies, this new understanding means a shift from episodic training interventions to continuous, self-organized learning ability. For investors, the paradigm of social learning design addresses the scaling possibilities of ‘culture-by-design’ development. For strategic partners, our model opens up a technological-social basis for the participatory implementation of complex innovation and transformation processes under conditions of uncertainty.
1. The Structural Problem: Why Psychological Safety does not Scale
- The evidence base
Amy Edmondson’s (1999) seminal work on psychological safety established its importance for team learning, performance and innovation. Google’s Project Aristotle (Rozovsky, 2015) confirmed these findings on a large scale and identified psychological safety as the most important factor in developing high-performance teams. Meta-analytic studies by Frazier et al. (2017) demonstrate the robust connection between psychological safety, learning behavior and organizational performance across different contexts. Nevertheless, two decades later, organizations are still struggling to systematically replicate these conditions beyond individual teams or charismatic leaders.
The reasons for this are primarily structural, not motivational. Research findings by Nembhard and Edmondson (2006) show that status differences, whether formal or informal, reliably suppress expression, even when managers explicitly ask for input. Milliken et al. (2003) demonstrate that employees systematically practice “voice withholding,” thereby creating organizational silence that persists despite proclaimed values of openness. In a comprehensive review, Morrison (2014) summarized that employees consciously withhold information, concerns and ideas, not out of disinterest, but out of fear of negative consequences, feelings of powerlessness or to protect relationships. Recent studies by Detert and Burris (2023) quantify the costs of this silence: organizations lose an estimated 25-40% of potential innovation impulses due to structurally induced self-censorship.
- The implementation gap
Current approaches to psychological safety fall into three categories, all of which have fundamental limitations:
(I) Cultural interventions focus on behavioral change among managers, workshops, coaches and normative appeals. While such interventions can certainly bring about local improvements, they remain dependent on leadership, culturally fragile and difficult to scale across organizational boundaries (Schein & Schein, 2017). A latest McKinsey analysis (2023) shows that 70% of transformation initiatives fail, primarily due to cultural barriers that such interventions cannot overcome.
(II) Measurement approaches (surveys, pulse checks, analytics) diagnose the problem but do not solve it. Knowing that a team lacks psychological safety does not make it safer to speak up (Delizonna, 2023).
(III) Platform solutions (collaboration tools, LMS) optimize content delivery or engagement metrics, for example, but leave social dynamics largely untouched. A Slack channel or discussion forum does not automatically reduce the social costs of admitting uncertainties or questioning consensus.
The remaining gap is fundamental: psychological safety is treated as an interpersonal climate that needs to be cultivated, rather than a system property that can be shaped or designed.
2. The Unsolved Corporate Challenge: Aligning a System’s Learning Conditions to its Social Architecture
2.1 Organizations as communication networks
Luhmann’s (1995) systems theory offers a crucial perspective: organizations are not arbitrary collections of people, but networks of communication that reproduce themselves through expectation-structures (autopoiesis). From this perspective, psychological safety is not about feeling comfortable, but about predictability and knowing what communication will follow one’s own contribution.
This changes the objective of the design task: instead of changing individual attitudes or behaviors, the task is to create conditions under which certain communication patterns among employees become more likely than others.
2.2 Learning under uncertainty
Argyris and Schön’s (1978) distinction between single-loop and double-loop learning remains fundamental. Single-loop learning optimizes within existing frameworks; double-loop learning questions the frameworks themselves. The latter requires admitting that current models of thinking may be inadequate which is a socially costly concession in most organizational contexts. Clark (2020) subsumed this dilemma in the fourth stage of his model of psychological safety at the level of challenger safety/ innovation safety.
Weick and Sutcliffe’s (2007) work on high-reliability organizations shows that adaptive systems institutionalize doubt under pressure, encourage diverse perspectives and draw on expertise regardless of rank. These are structural characteristics, not cultural aspirations.
2.3 The social costs of learning
2.3.1 Reputation management and interpretive authority
Goffman’s (1959) dramaturgical framework illuminates why learning is inherently risky: admitting ignorance, asking “obvious” questions, expressing justified criticism or proposing contradictory ideas jeopardizes the socially constructed self (the public persona) in hierarchies, which managers in particular maintain for themselves and others. In knowledge work, where expertise, reputation, interpretive authority and, directly linked to this, decision-making power over financial resources are hard currency, these costs are particularly high.
Kegan and Lahey’s (2009) Immunity to Change model demonstrates that organizational learning failure is often a rational response to competing self-constructs. The need to appear competent conflicts with the need to learn or to visibly signal identified competence requirements. Without addressing this structural tension, appeals to psychological vulnerability remain ineffective.
2.3.2 Social Load as a feature of modern working environments
Social load, our internal description for the social costs of organizational learning (not to be confused with cognitive load) refers to the cumulative psychological effort of social interactions in highly technical working environments. It manifests itself particularly where people have to maintain multiple social roles, communication channels and coordination logics in parallel, which is a characteristic of modern, highly networked organizations (Baecker, 2023).
A key driver of social load is multitasking in socially mediated work, which not only generates cognitive switching costs but also requires emotional and relational adjustment. Empirical studies show that frequent interruptions and parallel task processing are subjectively perceived as efficient, but in fact increase stress and reduce the quality of work and decisions (Mark et al., 2008).
In addition, there are specific friction losses in the interaction between human and AI-based communication. These arise, among other things, from different expectations, response logics and attributions of responsibility, as well as from the lack of shared situational contexts. Such human-machine asymmetries increase the coordination effort and shift additional (meta)cognitive and affective work to the human side (Suchman, 2007; Turkle, 2011).
Another significant factor of social load is continuous on-boarding in dynamic project and network structures: the ongoing adjustment to new team members, changing roles, implicit rules, social micro-norms and, not infrequently, the micro-dramas associated with them (Heinze, 2024). This form of relational and emotional work often remains invisible in classic productivity and efficiency models, but it requires substantial attention and emotional energy (Ries & Scholz, 2017).
In addition, social load encompasses the communicative stress of networking in organizational ecosystems that expect permanent visibility, connectivity and relationship building. Unlike occasional communication requirements, this constant exposure creates a latent burden, as social presence, responsiveness and self-positioning must be kept active even beyond clearly defined work roles.
Social load acts as a hidden dimension of stress that influences learning ability, decision-making quality, innovative strength and psychological safety, especially when people have to constantly calculate whether they can speak or not (“Can I ask this question? Will I appear incompetent?”). This invisible self-censorship costs performance (Edmondson, 2018).
Since social load was defined broadly in this initial presentation, both as a theoretical category and as an operational metric, we advocate to further investigate this crucial, yet hitherto neglected factor of a modern pluralistic workforce as one of our top-topics at NextGen.LX.
Joana Stella Kompa 