Fairness and governance of dual-track scoring on digital competition platforms: evidence from 34 seasons of Dancing with the Stars
Anzhe ChuPurpose
Dual-track entrepreneurial platforms integrate expert and public scores, but confidential voting figures create a data black box when evaluations from these two sources conflict. This study aims to examine how alternative mechanisms for integrating scores from these different sources shape screening fairness and how governance design can balance fairness with platform vitality.
Design/methodology/approach
Using 34 seasons of Dancing with the Stars as a proxy sample that is structurally analogous to dual-track entrepreneurial platforms that integrate expert assessments and public evaluations in screening decisions, this study applies inverse optimization, ordinal logistic regression and cardinal convex optimization to reconstruct unobserved public-preference trajectories. A third-order information-entropy variation system is then used to compare the institutional efficiency of alternative score-aggregation mechanisms.
Findings
Changes in scoring rules alter the elimination outcomes in 35% of borderline and controversial cases, while extremely low expert scores trigger a sympathy-vote rebound in subsequent stages. The proposed dynamic-weighting and dual-track-elimination mechanism gives greater weight to public input in early stages and shifts toward expert input in later stages. Historical backtesting raises the fairness score from 0.620 to 0.712 while maintaining platform vitality.
Research limitations/implications
The use of entertainment data limits generalizability because entertainment competitions and entrepreneurial ventures employ different evaluation criteria. Future research should validate the findings using commercial crowdfunding data.
Practical implications
The proposed mechanism helps platform designers balance public participation with professional risk control when configuring scoring, screening and elimination rules.
Originality/value
This study combines inverse reconstruction with an entropy-based framework to evaluate algorithmic fairness when proprietary platform data are unavailable, offering a governance tool for dual-track entrepreneurial ecosystems.