NeuroIS Literature Review

The field in one view

A small but institutionalized subfield with a mature methods conversation—and a striking shortage of neural work on generative AI and interactive agents.

The strongest dissertation opportunity sits at an intersection.

Dual-process models of anthropomorphism distinguish a fast, implicit response from a slower, reflective one. Yet those processes have been measured mainly with priming and self-report. At the same time, mental-state attribution toward text-based generative AI remains nearly unstudied with neural methods. Time-resolved EEG can test both problems in one coherent program.

10major research streams

5dissertation-ready gaps

2009first Gmunden Retreat

2015Springer proceedings begin

Important citation corrections

The canonical 2011 commentary by Dimoka, Pavlou & Davis appeared in Information Systems Research, not MISQ. Riedl, Davis & Hevner (2014) appeared in JAIS, not EJIS. “On the Foundations of NeuroIS” (2010) appeared in CAIS, not a Springer volume.

EEG / ERPtiming, load, error detection

fMRIlocalization, trust, ToM

Eye trackingattention, search, usability

EDA / HRVarousal and stress

fNIRSportable cortical activity

CortisolHPA stress response

Ten research streams

Select a stream to see its methods, constructs, landmark work and unresolved limitations.

Methods are only as good as their construct mapping

The strongest mappings have converging support outside IS. The weakest turn a generic physiological response into a specific psychological claim.

Construct

Best-supported measures

What the signal actually supports

Validity

Cognitive load

Frontal theta ↑; parietal alpha ↓; P300/P200; pupil dilation

Workload and attentional-resource demand, with task and luminance controls

Strong

Stress / arousal

EDA/SCR; HRV; cortisol

Sympathetic activation, autonomic balance and HPA response—not a named emotion

Strong

Attention

Fixations/saccades; P300; alpha suppression

Where and when resources are allocated; fixation does not equal preference

Strong

Error monitoring

ERN / ErrP / FRN

Detection of mismatch or unfavorable feedback; useful for robot and AI errors

Strong

Mental-state attribution

mPFC / TPJ / precuneus; N170 and LPP timing

ToM-network engagement and stages of face/intent processing; context still matters

Moderate–strong

Trust / distrust

fMRI contrasts; reliance behavior; self-report

No biosignal directly measures trust. Triangulation is essential.

Caution

Valence / engagement

Facial EMG; frontal alpha asymmetry; vendor indices

Facial EMG can support valence; proprietary “engagement” scores are often opaque

Mixed

Reverse inference is the central validity threat.

Activation in a region does not entail a particular mental state because most regions serve multiple functions. Physiological signals should be paired with behavior and self-report; when the measures diverge, that divergence may be the theoretically interesting result. See Poldrack’s review and the JAIS methodology agenda.

Five dissertation-level gaps

Each gap pairs a theoretical hole with a measurement problem and a specific NeuroIS design. The first three best match the proposed thesis direction.

Read these next

A focused 2024–2026 list, ordered to move from the core thesis question to design and methods. “Provisional” means preprint or press-level evidence.

Where design science fits

The most feasible pairing is straightforward: build a trust-calibration intervention with standard DSRM, then evaluate whether it restores neural and behavioral verification—not merely whether users say they like it.

ProblemObjectivesDesign & buildDemonstrateEvaluateCommunicate

Neuroscience informs the design

Use dual-process anthropomorphism and cognitive-load knowledge to derive design principles for uncertainty cues, verification friction or AI-agent embodiment. This path can use neuroscience knowledge without collecting biosignals.

NeuroIS evaluates the artifact

Compare a calibration interface against a baseline with ERN/FRN, P300, frontal theta, eye tracking and reliance behavior. This is the strongest near-term dissertation pairing.

Neurophysiology becomes the artifact

Build a neuroadaptive agent that detects error awareness, workload or engagement and changes its behavior in response. NeuroChat is an early template; passive-BCI ErrP work supplies the control signal.

Evaluation rule

Report reliability, validity, sensitivity, diagnosticity, objectivity and intrusiveness for every physiological measure. A DSR artifact does not relax the NeuroIS validity burden.

What this review can—and cannot—claim

Strong research positioning depends on being explicit about where the evidence thins out.

Methodological pattern

Landmark fMRI and HRI studies often use small samples: the reviewed examples range from n=6 pilots to roughly n=42.
Most studies are controlled lab experiments with student samples and simplified tasks.
Multimodal triangulation is advocated more often than it is delivered.
No registered replication of a landmark NeuroIS result was located in this review.
Task-fMRI reliability problems in the wider neuroscience literature make single-study localization especially fragile.

Coverage gaps

This is a source-backed thematic synthesis, not a PRISMA meta-analysis with database-wide inclusion counts.
The WI/Wirtschaftsinformatik track was not separately surveyed.
fNIRS and facial EMG remain thin in core IS venues.
Flow appears in Retreat calls but no landmark study was identified with confidence.
Algorithmic management plus neurophysiology produced no empirical match; AI-mediated communication produced one 2026 study.
Two recent items are provisional and should not anchor a citation until their scholarly record is confirmed.

Selected source trail

Direct links to the field-defining sources and evidence most important to the synthesis.

Dimoka, Pavlou & Davis (2011). “Research Commentary—NeuroIS.” Information Systems Research, 22(4), 687–702.

Riedl et al. (2010). “On the Foundations of NeuroIS.” CAIS, 27, 243–264.

Dimoka et al. (2012). “On the Use of Neurophysiological Tools in IS Research.” MIS Quarterly, 36(3), 679–702.

Riedl, Davis & Hevner (2014). “Towards a NeuroIS Research Methodology.” JAIS, 15(10).

Riedl & Léger (2015/2016). Fundamentals of NeuroIS. Springer.

Dimoka (2010). “What Does the Brain Tell Us About Trust and Distrust?” MIS Quarterly, 34(2), 373–396.

Dumont et al. (2018). “Non-invasive Brain Stimulation in IS Research.” PLoS ONE, 13(7).

Krach et al. (2008). “Can Machines Think?” PLoS ONE, 3(7).

Spatola & Chaminade (2022). Precuneus response in human–robot versus human–human interaction. Scientific Reports.

Złotowski et al. (2018). “Model of Dual Anthropomorphism.” International Journal of Social Robotics, 10, 701–714.

Passive BCI for Enhanced HRI (2020). Frontiers in Robotics and AI.

Paul, Turner & Miller (2018). Evidence on task-fMRI replicability. Communications Biology.