Research opportunity · What kind of thing is generative AI, and does the answer predict anything?
When AI output goes wrong at work, who answers for it?
In short: has this been done?
Related work: Partly addressed. Related studies cover parts of this question; what they leave open is set out below. Only a first-pass search has been done, so it is not yet known whether a fuller search would find a study that does the whole design.
How far checked: First-pass screen only. A first-pass AI web search on 16 October 2026. No deeper check yet. Every search and check so far was done by AI; no person or domain expert has checked them yet.
The searches, the nearest work found and what the search could not reach
This page shows RO-LD01-04 version v1, whose record never changes. It is the current version; if the opportunity is revised, the newer version appears at its main page.
Question
In organizations that use generative AI, how is responsibility for an AI-produced error assigned in practice (to the employee who used the tool, the manager, a risk team, the vendor or no one), and does the assignment differ when the system is described as a tool, an assistant or an agent?
Gap type: Institutional practice unknownA PhD projectLow to medium costRelated work: Partly addressed
Why it matters
The artificial-communication account places its observables on the human side, in who is held accountable for outputs and how they gain standing, but the sources read for this debate do not measure that side. The agent account and agency law make the same question practical: the Restatement (Third) of Agency, as one legal scholar quotes it, treats programs as instrumentalities of their users "at present", while firms market agents. How organizations actually assign responsibility would show which description is in use where it matters.
The debate lists this among what we still do not know: What does the artificial-communication account predict about accountability and uptake, and can it be measured?
Evidence that motivates it
- Esposito argues that comparing algorithms with human intelligence is misleading and proposes thinking of them “not in terms of artificial intelligence but in terms of artificial communication”: if machines contribute to social intelligence, “it will not be because they have learned how to think like us but because we have learned how to communicate with them.” (LD01-SL-C08 v1.0)
How this study uses it: Machines contribute to social intelligence because people learn to communicate with them. - Drawing on Luhmann, Esposito, Wittgenstein and Brandom, Fenoglio argues that what language models lack is a position in the practices of commitment and assessment, not a behavioural capacity, so that “no behavioral evidence can satisfy a positional criterion, and this is by design rather than by evasion.” (LD01-SL-C09 v1.0)
How this study uses it: One reading places what models lack in practices of commitment and assessment, not behaviour. - Kolt notes that the US Restatement (Third) of Agency holds that a computer program “is not capable of acting as a principal or an agent as defined by the common law” and that “at present, computer programs are instrumentalities” of their users, and he uses agency law and principal–agent economics to analyse AI agents’ problems of information asymmetry, discretionary authority and loyalty. (LD01-SL-C11 v1.0)
How this study uses it: The Restatement (Third) of Agency, as quoted by the author, treats a computer program as an instrumentality of its user "at present"; the author calls this the apparent consensus and notes contrary views. - Anthropic distinguishes “workflows,” in which language models and tools are orchestrated through predefined code paths, from “agents,” in which models “dynamically direct their own processes and tool usage”; it says agents are “typically just LLMs using tools based on environmental feedback in a loop.” (LD01-SL-C13 v1.0)
How this study uses it: A developer defines agents by control flow and markets them as such.
Each claim is shown as it read when this version was published.
A possible design
Combine an organizational study with an experiment. Collect incident records and interviews in a sample of firms in two or three sectors to trace how AI-related errors were attributed, escalated and resolved. Alongside, run a pre-registered vignette experiment with managers that varies how the system is described (tool, assistant, agent) and its degree of autonomy, measuring blame and remedies. Compare the field and experimental patterns.
Data needed
Firm access for incident records and interviews; a manager panel for the experiment; coding scheme for attribution.
Feasibility
A PhD project. Low to medium; panel fees and fieldwork time. Firms may not share incident records, so interviews may carry most of the field evidence. Surveys reported in the press in 2026 give starting points but were not read for this debate.
Has this been done?
Partly done already: related work covers some of this question (first-pass screen of 16 October 2026; not yet checked in depth).
Surveys and small vignette studies of blame for AI errors exist, but no study was found that traces attribution in organizations' actual incidents or links it to how the system is described.
How strong this evidence is
- Who searched and checked: no checker is recorded beyond the first-pass screen, which was part of the AI drafting (Claude (SocioLens Project Lead), AI-drafted from the LD-01 debate). Every check so far was done by AI; no person or domain expert has checked it yet.
- What was read: 2 of the 2 nearest works below were judged from an abstract, summary or search listing only; each is marked. Not every entry says how much of the work was read.
- Where the search could not reach: Screened by one web search and the LD-01 source scan on the date shown; neither nearest work was read; management and law journals were not searched. Absence of a hit is not proof that the work does not exist.
- Before building a project on it, run your own search in a scholarly database such as Scopus, Web of Science, Google Scholar or SSRN.
Nearest work found
- Wharton School survey of managers on accountability for AI errors (October 2026), as reported in the press Judged from an abstract, summary or search listing only
Seen in a press report only; the paper was not read. Reports managers split between the employee, the manager and the risk team. - Vignette study of responsibility in human–AI collaboration, AMCIS 2026 Judged from an abstract, summary or search listing only
Seen in search results only. A small experiment (reported N = 48) on how beliefs about automation shift blame.
Search queries used
who is held accountable for errors in AI-generated outputs in organizations responsibility attribution generative AI workplace empiricalLD-01 source scan for studies of accountability under the artificial-communication account (16 October 2026)blame attribution AI described as agent versus tool experiment managers
Cite this page
SocioLens project. (2026, October 16). When AI output goes wrong at work, who answers for it? [Research opportunity RO-LD01-04, version v1]. SocioLens. https://sociolens-web.pages.dev/opportunities/ro-ld01-04/v1/
@misc{sociolens_ro_ld01_04_v1,
author = {{SocioLens project}},
title = {When AI output goes wrong at work, who answers for it?},
year = {2026},
date = {2026-10-16},
howpublished = {\url{https://sociolens-web.pages.dev/opportunities/ro-ld01-04/v1/}},
note = {Research opportunity RO-LD01-04, version v1}
}The link points to this version, whose record never changes. For a finding, cite the original study as well: SocioLens summarizes research and does not replace it.
Provenance
RO-LD01-04 v1. Drafted by Claude (SocioLens Project Lead), AI-drafted from the LD-01 debate on 2026-10-16. Not yet reviewed. Not yet reviewed by a domain expert. This is a research hypothesis, not a finding.