Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Characterizing Human Actions in the Digital Platform by Temporal Context

Akira Matsui, Emilio Ferrara

TL;DR

The paper addresses the missing inter-temporal context in action logs from digital platforms. It introduces the Action Timing Context (ATC) framework, which jointly embeds each action with its time interval by discretizing intervals through a mixture of exponential distributions, constructing action sequences with time-bin tokens, and learning embeddings via Skip-gram with negative sampling. By comparing long-term and short-term reference contexts, ATC assigns each action a context label that reflects underlying cognitive states and pacing. Empirical analyses across app usage, MOOC, and StudentLife datasets demonstrate that ATC captures meaningful variations in behavior, differences between dropout and non-dropout participants, and dynamic transitions over time, yielding a unified and interpretable view of inter-temporal human activity on digital platforms.

Abstract

Recent advances in digital platforms generate rich, high-dimensional logs of human behavior, and machine learning models have helped social scientists explain knowledge accumulation, communication, and information diffusion. Such models, however, almost always treat behavior as sequences of actions, abstracting the inter-temporal information among actions. To close this gap, we introduce a two-scale Action-Timing Context(ATC) framework that jointly embeds each action and its time interval. ATC obtains low-dimensional representations of actions and characterizes them with inter-temporal information. We provide three applications of ATC to real-world datasets and demonstrate that the method offers a unified view of human behavior. The presented qualitative findings demonstrate that explicitly modeling inter-temporal context is essential for a comprehensive, interpretable understanding of human activity on digital platforms.

Characterizing Human Actions in the Digital Platform by Temporal Context

TL;DR

The paper addresses the missing inter-temporal context in action logs from digital platforms. It introduces the Action Timing Context (ATC) framework, which jointly embeds each action with its time interval by discretizing intervals through a mixture of exponential distributions, constructing action sequences with time-bin tokens, and learning embeddings via Skip-gram with negative sampling. By comparing long-term and short-term reference contexts, ATC assigns each action a context label that reflects underlying cognitive states and pacing. Empirical analyses across app usage, MOOC, and StudentLife datasets demonstrate that ATC captures meaningful variations in behavior, differences between dropout and non-dropout participants, and dynamic transitions over time, yielding a unified and interpretable view of inter-temporal human activity on digital platforms.

Abstract

Recent advances in digital platforms generate rich, high-dimensional logs of human behavior, and machine learning models have helped social scientists explain knowledge accumulation, communication, and information diffusion. Such models, however, almost always treat behavior as sequences of actions, abstracting the inter-temporal information among actions. To close this gap, we introduce a two-scale Action-Timing Context(ATC) framework that jointly embeds each action and its time interval. ATC obtains low-dimensional representations of actions and characterizes them with inter-temporal information. We provide three applications of ATC to real-world datasets and demonstrate that the method offers a unified view of human behavior. The presented qualitative findings demonstrate that explicitly modeling inter-temporal context is essential for a comprehensive, interpretable understanding of human activity on digital platforms.
Paper Structure (27 sections, 8 equations, 7 figures, 2 tables)

This paper contains 27 sections, 8 equations, 7 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Research and Theoretical Background
  3. Inter‑temporal information, actions, and human cognitive states
  4. User modeling with word embedding
  5. Method
  6. Capturing inter‑temporal information with time bins
  7. Time‑interval‑bin estimation by a mixture of exponential distributions
  8. Constructing action n‑gram sequences
  9. Action sequence
  10. Action n‑gram
  11. Embedding action n‑gram sequences
  12. "Word’’ and "Context’’ in this paper
  13. Extracting action‑timing context (ATC) using n‑gram actions
  14. Constructing the reference vectors
  15. Definition of "long‑term context’’ and "short‑term context’’
  16. ...and 12 more sections

Figures (7)

  • Figure 1: Schematic of studying the inter-temporal context of user actions
  • Figure 2: Extracting action‑timing contexts with n‑gram actions
  • Figure 3: Smartphone Apps usage context
  • Figure 4: Action timing context differences: Drop-out students VS Non-dropout students
  • Figure 5: Transition of the action timing context of the students over the academic term
  • ...and 2 more figures