Mediation Challenges and Socio-Technical Gaps for Explainable Deep Learning Applications
Pith reviewed 2026-05-24 20:52 UTC · model grok-4.3
The pith
Deep learning experts expect mediators to connect technical model work with social meanings but current incentives discourage that work.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
Participating DL experts did not spontaneously engage with the social meaning of the models they build; when stimulated they expected mediators to bridge technical and social meanings; current incentives guiding their work are at odds with the scientific challenges of advancing XAI and responding to societal demands for explanations.
What carries the argument
XAI Mediation Challenges: three preliminary challenges proposed to bring together technical and social meanings of DL applications and foster interdisciplinary collaboration.
If this is right
- Without changes in incentives, progress on XAI that satisfies societal demands such as the right to explanation will remain limited.
- Interdisciplinary collaboration between AI researchers and social scientists is required to address the identified gaps.
- The three proposed mediation challenges offer concrete starting points for such collaboration.
- Real-world DL applications will need dedicated mediators if technical and social meanings are to be aligned.
Where Pith is reading between the lines
- Similar incentive misalignments may exist in other subfields of AI that face external regulatory or ethical pressures.
- Labs could experiment with internal roles or review processes that reward attention to social meaning without waiting for field-wide incentive reform.
- The mediation challenges could be tested in pilot projects where social scientists are embedded in DL teams from the start.
Load-bearing premise
Findings from one group of DL experts in a single industrial lab can be generalized to research incentives across the broader XAI field.
What would settle it
A larger multi-lab survey or ethnographic study that finds DL researchers routinely and spontaneously incorporate social-meaning considerations into their daily work would undermine the central claim.
read the original abstract
The presumed data owners' right to explanations brought about by the General Data Protection Regulation in Europe has shed light on the social challenges of explainable artificial intelligence (XAI). In this paper, we present a case study with Deep Learning (DL) experts from a research and development laboratory focused on the delivery of industrial-strength AI technologies. Our aim was to investigate the social meaning (i.e. meaning to others) that DL experts assign to what they do, given a richly contextualized and familiar domain of application. Using qualitative research techniques to collect and analyze empirical data, our study has shown that participating DL experts did not spontaneously engage into considerations about the social meaning of machine learning models that they build. Moreover, when explicitly stimulated to do so, these experts expressed expectations that, with real-world DL application, there will be available mediators to bridge the gap between technical meanings that drive DL work, and social meanings that AI technology users assign to it. We concluded that current research incentives and values guiding the participants' scientific interests and conduct are at odds with those required to face some of the scientific challenges involved in advancing XAI, and thus responding to the alleged data owners' right to explanations or similar societal demands emerging from current debates. As a concrete contribution to mitigate what seems to be a more general problem, we propose three preliminary XAI Mediation Challenges with the potential to bring together technical and social meanings of DL applications, as well as to foster much needed interdisciplinary collaboration among AI and the Social Sciences researchers.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper reports a qualitative case study with Deep Learning experts from one industrial R&D laboratory. Participants did not spontaneously discuss the social meaning of their models; when prompted, they expected mediators to bridge technical and social interpretations in deployed applications. The authors conclude that prevailing research incentives conflict with the requirements for advancing XAI to meet societal demands such as GDPR rights to explanation, and they propose three preliminary XAI Mediation Challenges to encourage interdisciplinary work between AI and social sciences.
Significance. If the observed patterns generalize, the study supplies practitioner-grounded evidence of socio-technical gaps in XAI and offers concrete mediation challenges that could help align technical development with user and societal needs while promoting cross-disciplinary collaboration. The empirical qualitative approach is a strength in providing direct observations from domain experts.
major comments (2)
- [Conclusion] Conclusion (final paragraph): The claim that 'current research incentives and values guiding the participants' scientific interests and conduct are at odds with those required to face some of the scientific challenges involved in advancing XAI' generalizes from a single industrial laboratory to field-wide structures. The manuscript supplies no comparative data from academic settings or other organizations, no direct measures of incentives, and no triangulation against alternatives such as role specialization or domain familiarity.
- [Methods] Methods section: The strength of the interpretive conclusions depends on the qualitative data collection and analysis procedures (interview protocol, participant sampling, coding approach). These details are referenced but not specified at a level that allows assessment of how the reported patterns (absence of spontaneous social-meaning discussion; expectation of mediators) were derived and whether they support the incentive-misalignment inference.
minor comments (2)
- [Abstract] Abstract: The three XAI Mediation Challenges are announced but neither named nor briefly characterized; adding one sentence of description would improve accessibility.
- [Introduction] Terminology: 'Social meaning' and 'mediators' are central but receive operational definitions only later; an early explicit definition would reduce ambiguity for readers from technical backgrounds.
Simulated Author's Rebuttal
We thank the referee for the detailed and constructive review. We address each major comment below, clarifying the scope of our case study and agreeing to strengthen the methods description where possible.
read point-by-point responses
-
Referee: [Conclusion] Conclusion (final paragraph): The claim that 'current research incentives and values guiding the participants' scientific interests and conduct are at odds with those required to face some of the scientific challenges involved in advancing XAI' generalizes from a single industrial laboratory to field-wide structures. The manuscript supplies no comparative data from academic settings or other organizations, no direct measures of incentives, and no triangulation against alternatives such as role specialization or domain familiarity.
Authors: We agree the conclusion draws an inference from a single-lab case study without comparative data or direct incentive measures. The claim is presented as an implication of the observed patterns (lack of spontaneous social-meaning discussion and expectation of mediators) rather than a proven field-wide fact. We will revise the final paragraph to explicitly frame the inference as context-specific to this industrial R&D setting and to note the absence of triangulation, while retaining the suggestion that the patterns may point to broader incentive issues warranting further study. revision: partial
-
Referee: [Methods] Methods section: The strength of the interpretive conclusions depends on the qualitative data collection and analysis procedures (interview protocol, participant sampling, coding approach). These details are referenced but not specified at a level that allows assessment of how the reported patterns (absence of spontaneous social-meaning discussion; expectation of mediators) were derived and whether they support the incentive-misalignment inference.
Authors: We accept that the current methods description is insufficiently detailed for independent assessment. The manuscript references the qualitative procedures but does not fully specify the interview protocol, sampling criteria, or coding process. We will expand the Methods section with these details (including how the absence of spontaneous discussion and mediator expectations were identified in the data) to support evaluation of the interpretive steps. revision: yes
Circularity Check
No circularity: empirical qualitative case study with conclusions drawn from interview data
full rationale
The paper is a qualitative case study reporting observations from interviews with DL experts in a single industrial lab. Its central claim—that research incentives conflict with XAI needs—is presented as an inference from the reported participant behavior (lack of spontaneous social-meaning discussion and expectation of mediators). No equations, derivations, fitted parameters, or self-referential modeling appear; the text contains no self-citation chains, uniqueness theorems, or ansatzes that reduce the conclusion to its own inputs by construction. The derivation chain is therefore self-contained as an empirical report rather than a closed logical loop.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.