Twin Papers: A Simple Framework of Causal Inference for Citations via Coupling (CIKM 2022)

We propose a simple framework for determining the best decision making in research processes. The basic idea is to regard a decision as an intervention and measure the outcome by the number of citations. However, it is difficult to test with actual interventions, e.g., by randomly changing publication venues or titles, because it may cause adverse impacts on the researchers' careers. In this paper, we are inspired by the existing counterfactual analysis using twins, where the researchers regard twins as counterfactual units, and we propose to regard a pair of papers that cite each other as twins.

Paper: https://arxiv.org/abs/2208.09862

💿 Dependency

$ pip install -r requirements.txt
$ sudo apt install wget unzip p7zip-full

🗃️ Download and Preprocess Datasets

You can download and preprocess data by the following command. It may take time and large memory.

$ bash download_and_preprocess.sh

Alternatively, you can download preprocessed data via

$ bash download_readymade.sh

• twin_ids.pickle lists twin papers.
• twin_papers.pickle stores various information of papers.
• self.pickle is a set of papers that are self-cited.
• coauthor.pickle is a collaboration network.

The definition of these files are described in extract_twins.py.

🧪 Experiments

Confirming Assumptions (Section 5)

The image above can be reproduced by the following command.

$ python assumptions.py

Analysis (Section 6)

Treatment	$|\mathcal{D}|$	$\widetilde{\text{ATE}}$
Including a Colon in the Title	21080	0.356
Lengthening the Title	84970	-0.126
Lengthening the reference	81857	0.710
Lengthening the abstract	82917	0.248
Lengthening the paper	65730	0.630
Self citation	10582	1.30
Earlier publication	41546	−0.187

The above table can be reproduced by analyze.py

Treatment (a)	Control (b)	$|\mathcal{D}^{a/b}|$	$\widetilde{\text{ATE}}$
Journal of Cognitive Neuroscience	NeuroImage	817	0.539
IEEE Transactions on Information Theory	International Symposium on Information Theory	459	1.93
Neural Computation	IEEE Transactions on Neural Networks	216	0.632
Neural Computation	Neural Networks	199	0.76
Symposium on the Theory of Computing	Foundations of Computer Science	182	0.252
IEEE Transactions on Signal Processing	International Conference on Acoustics, Speech, and Signal Processing	178	2.78
Neural Computation	Neurocomputing	153	2.41
Journal of Economic Theory	Games and Economic Behavior	125	0.668
IEEE ACM Transactions on Networking	International Conference on Computer Communications	110	1.29
Symposium on the Theory of Computing	Symposium on Discrete Algorithms	105	0.693

The above table can be reproduced by venue.py

🖋️ Citation

@inproceedings{sato2022twin,
  author    = {Ryoma Sato, Makoto Yamada, Hisashi Kashima},
  title     = {Twin Papers: A Simple Framework of Causal Inference for Citations via Coupling},
  booktitle = {The 31st {ACM} International Conference on Information
               and Knowledge Management, {CIKM}},
  year      = {2022},
}