Incentivized Network Dynamics in Digital Job Recruitment

TL;DR

The paper tackles recruiting passive candidates by introducing the Independent Halting Cascade (IHC), an agent-based diffusion–halting model that ties information spread to actionable job applications. It extends the Independent Cascade framework with incentives parameterized by $\beta$ and skill–vacancy matching, generating diffusion, halting, and failure regimes across ER, BA, and homophilic networks. The authors derive analytical boundaries (e.g., $\langle k \rangle p_r p_a p_h = 1$ for diffusion) and validate them via simulations that reproduce Travers–Milgram and Dodds chain-length distributions, while demonstrating robust performance on real networks with fewer applicants than direct-recommendation baselines. The framework offers theoretical insights and practical guidance for designing recruitment systems that effectively engage passive candidates and coordinate task completion in networked settings.

Abstract

Recruiting passive candidates, i.e., individuals not actively seeking jobs but open to compelling opportunities, remains one of the hardest challenges in digital recruitment. Motivated by a real collaboration with an industry partner, we introduce the Independent Halting Cascade (IHC) model: a simple but rich agent-based framework that couples network diffusion with the possibility of halting through job applications. Agents can either recommend vacancies to peers or apply themselves, and incentives increase the likelihood of recommendation, mobilizing otherwise passive candidates. The IHC bridges research on social network diffusion, coordinated task completion, and labor economics by modeling heterogeneous skills, job specificities, and network structures, including homophily. We derive analytical boundaries that characterize diffusion and failure regimes, and we show, through simulations, that the IHC reproduces the empirical chain-length distributions of Travers and Milgram, and of Dodds, with only coarse calibration. Across synthetic (ER, BA, homophilic) and real networks (SMS, e-mail, Twitter), the IHC achieves comparable or higher success rates than direct-recommendation baselines, while requiring fewer applicants. Our findings suggest that the IHC captures core mechanisms of coordinated task completion, offering both a theoretical contribution and a practical foundation for recruitment systems designed to reach and engage passive candidates.

Figures (10)

• Figure 1: Independent Halting Cascade (IHC) model diagram: The initial spreader(s), $u_0$ (blue), recommends its neighbors, $u_1$, $u_2$, and $u_3$, with probability $p_r(u_0, u_1)$, $p_r(u_0, u_2)$, and $p_r(u_0, u_3)$, respectively. Active recommenders, here $u_1$ (green), try to recommend their neighbors with probability of $1 - p_a(u_1)$. Otherwise, they apply to halt the chain and succeed with probability $p_h(u_1)$. The cascade continues until someone successfully halts the chain (a successful chain) or there are no new recommendations (an unsuccessful chain).
• Figure 2: IHC model behavior on Erdős–Rényi (ER) networks with homogeneous parameters. Average chain length (top row), number of applicants (middle row), and success rate (bottom row) as functions of recommendation probability $p_r$ and application probability $p_a$, for three levels of hiring probability $p_h$ ($0.1$, $0.5$, and $1$, from left to right). The solid black line marks the diffusion boundary, where most cascades reduce to direct-recommendation chains. The dashed black line marks the failure boundary, below which cascades are more likely to die out without a successful halt. As $p_h$ decreases, the direct-recommendation region shrinks, and longer chains sustained by social recommendations become crucial for success.
• Figure 3: Effect of increasing incentives on IHC dynamics.A. Analytic transformation of the diffusion boundary under the incentive function (Eq. \ref{['eq:incentives_pr']}). As the incentive strength $\beta$ increases, the boundary shifts downward, reducing the baseline recommendation probability $p_r$ required for cascades to succeed. B. Success rate and C. average chain length from simulations on Erdős–Rényi networks with $N = 2000$, $\langle k \rangle = 20$, $p_h = 0.01$, and $p_a = 0.7$. Increasing incentives drive the success rate to one even at small $p_r$ values, while simultaneously shrinking the region of long cascades.
• Figure 4: Effect of the initial spreader’s connectivity on IHC dynamics in BA and ER networks.A. Success rate and B. number of applicants as functions of the recommendation probability $p_r$ for Erdős–Rényi (ER) networks with $\langle k \rangle = 20$ (red, solid) and Barabási–Albert (BA) networks conditioned on the initial spreader’s degree $k_0$ (shaded lines), with $p_h = 0.1$ and $p_a = 0.25$. Higher $k_0$ values in BA networks increase both the success rate and the number of applicants.
• Figure 5: Comparison of IHC (red) and direct-recommendation systems (blue) in ER networks.A. Success rate, B. average chain length, and C. average number of applicants as functions of the recommendation probability $p_r$ for vacancies with $n_\nu \in \{4,6,8\}$ requirements. The direct recommendation system has access to half of the population ($\rho = 0.5$), always produces chains of length two, and yields a linear increase in applicants with $p_r$. The IHC achieves comparable success rates even for specific vacancies, at the cost of longer but more efficient chains with fewer applicants.

Theorems & Definitions (2)

• Proposition 1: Direct-halting boundary
• Proposition 2: Failure boundary