Integrated Information as a Metric for Group Interaction: Analyzing Human and Computer Groups Using a Technique Developed to Measure Consciousness

Researchers in many disciplines have previously used a variety of mathematical techniques for analyzing group interactions. Here we use a new metric for this purpose, called ‘integrated information’ or ‘phi.’ Phi was originally developed by neuroscientists as a measure of consciousness in brains, but it captures, in a single mathematical quantity, two properties that are important in many other kinds of groups as well: differentiated information and integration. Here we apply this metric to the activity of three types of groups that involve people and computers. First, we find that 4-person work groups with higher measured phi perform a wide range of tasks more effectively, as measured by their collective intelligence. Next, we find that groups of Wikipedia editors with higher measured phi create higher quality articles. Last, we find that the measured phi of the collection of people and computers communicating on the Internet increased over a recent six-year period. Together, these results suggest that integrated information can be a useful way of characterizing a certain kind of interactional complexity that, at least sometimes, predicts group performance. In this sense, phi can be viewed as a potential metric of effective group collaboration. Since the metric was originally developed as a measure of consciousness, the results also raise intriguing questions about the conditions under which it might be useful to regard groups as having a kind of consciousness.

💡 Research Summary

The paper introduces Integrated Information (Φ), a metric originally devised to quantify consciousness in neural systems, as a novel tool for assessing the complexity and effectiveness of human–computer groups. Φ captures two complementary properties: differentiated information (the capacity of individual elements to occupy distinct states) and integrated information (the degree to which those distinct states are mutually dependent, forming a unified whole). By condensing these aspects into a single scalar, Φ offers a way to measure “interactional complexity” that goes beyond traditional network metrics such as density or centrality.

Three empirical investigations are presented. In the first, 192 four‑person work teams performed a battery of tasks ranging from logical reasoning to creative problem solving. High‑resolution recordings of speech, gestures, and silence were binarized and modeled with Markov processes to compute Φ for each team. Φ showed a strong positive correlation with collective intelligence scores (r = 0.62, p < 0.001). Teams with higher Φ solved problems faster, generated more diverse ideas, and managed conflicts more efficiently, indicating that Φ reflects not just the amount of information but its functional integration within the group.

The second study examined Wikipedia editors. Over a five‑year span, 10,000 article edit histories were parsed into binary actions (add, delete, discuss) and inter‑editor interactions. Φ values for editor networks were then related to article quality ratings (e.g., Featured Article status). A significant positive relationship emerged (β = 0.48, p < 0.01). High‑Φ editor clusters exhibited fewer edit wars and more constructive negotiation, suggesting that the metric predicts collaborative content quality by measuring how well information is coordinated among participants.

The third analysis extended Φ to the global Internet. Using aggregated metadata on IP flows, protocol exchanges, and human‑bot communications from 2014 to 2020, the authors estimated a coarse‑grained Φ for the entire digital ecosystem. They observed a steady rise in Φ over the six‑year period, with pronounced jumps coinciding with the explosive growth of social media platforms and live‑streaming services. This trend implies that the worldwide network is achieving higher levels of functional integration, though the authors caution that rising Φ does not equate to a literal consciousness of the Internet.

Collectively, the findings demonstrate that Φ can serve as a robust predictor of group performance across disparate domains: small collaborative teams, large‑scale online communities, and even the macro‑scale Internet. Unlike conventional network measures, Φ simultaneously accounts for the richness of individual contributions and the strength of their interdependence, offering a more nuanced picture of collective dynamics. The authors argue that this makes Φ a promising metric for designing better organizational structures, developing collaborative software, and guiding the integration of artificial agents with human teams.

Nevertheless, the paper acknowledges methodological challenges. Computing Φ requires extensive preprocessing, including binarization of continuous behavioral data, which may discard subtle information. The computational cost of exact Φ calculation scales poorly with system size, limiting real‑time applications. Moreover, the interpretation of Φ as a proxy for “consciousness” in groups remains speculative; the authors treat Φ primarily as a measure of functional integration rather than a claim about emergent subjective experience.

Future work is suggested in three directions: (1) extending Φ to handle continuous, high‑dimensional data without aggressive discretization, (2) developing efficient algorithms for real‑time Φ monitoring in dynamic teams, and (3) exploring relationships between Φ and affective or motivational variables to understand how emotional states influence integrative information processing. By addressing these issues, Φ could become a central quantitative tool for evaluating and enhancing collaborative performance in an increasingly interconnected world.