arxiv: 2604.19489 · v1 · submitted 2026-04-21 · 💻 cs.CV · cs.CY

Recognition: unknown

Seeing Candidates at Scale: Multimodal LLMs for Visual Political Communication on Instagram

Michael Achmann-Denkler , Mario Haim , Christian Wolff

Authors on Pith no claims yet

Pith reviewed 2026-05-10 02:32 UTC · model grok-4.3

classification 💻 cs.CV cs.CY

keywords multimodal LLMsvisual political communicationInstagramface recognitionperson countingpolitical campaignsGPT-4oelection analysis

0 comments

The pith

GPT-4o outperforms traditional computer vision models at recognizing politicians and counting people in Instagram campaign images.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper compares specialized machine learning models with multimodal large language models for analyzing visual political communication on Instagram. It examines images from the 2021 German federal election to identify front-runner politicians and count individuals in stories and posts. GPT-4o achieved macro F1-scores of 0.89 for face recognition and 0.86 for person counting, surpassing FaceNet512, RetinaFace, and Google Cloud Vision. These results indicate that advanced AI can process large volumes of social media visuals that influence public views of candidates. Scaling such analysis matters because manual review of campaign imagery is too slow for comprehensive studies.

Core claim

The paper shows that the multimodal large language model GPT-4o outperforms traditional computer vision models in identifying front-runner politicians and counting individuals in Instagram stories and posts during the 2021 German federal election campaign, reaching a macro F1-score of 0.89 for face recognition and 0.86 for person counting while highlighting the potential of such systems to scale visual political communication analysis.

What carries the argument

Multimodal large language model GPT-4o applied to simultaneous face recognition of politicians and person counting in campaign imagery from Instagram.

Load-bearing premise

The manually created ground-truth labels for politician identities and person counts are accurate and free of systematic bias across the Instagram dataset.

What would settle it

Independent re-annotation of a random sample of the Instagram images by multiple human coders, where low inter-annotator agreement with the original labels would show the evaluation metrics rest on flawed ground truth.

Figures

Figures reproduced from arXiv: 2604.19489 by Christian Wolff, Mario Haim, Michael Achmann-Denkler.

**Figure 2.** Figure 2: To measure concentrated visibility, we use the [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 3.** Figure 3: The prompt used to determine the presence of a political leader in social media images. The [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

**Figure 4.** Figure 4: The face recognition workflow starts with face [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗

**Figure 6.** Figure 6: We filtered the corpus to include only images [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗

**Figure 7.** Figure 7: Confusion matrices comparing the computational classifications (predicted labels) with the human annotations [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗

**Figure 8.** Figure 8: Several examples of stories where the model [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗

**Figure 9.** Figure 9: Three confusion matrices comparing the person count performance across the four labels and between the [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗

**Figure 10.** Figure 10: Three examples of images where the RetinaFace model detected faces. In the face recognition stage, human annotators identify the faces. When counting persons, human annotators agreed that the correct person count is 0. 4 Results: Case Study In this section, we present the findings from our case study analysis of concentrated visibility in visual political communication. The focus is on identifying patte… view at source ↗

**Figure 11.** Figure 11: Comparison of front-runner appearances in Instagram [PITH_FULL_IMAGE:figures/full_fig_p014_11.png] view at source ↗

**Figure 12.** Figure 12: Comparison of front-runner appearances in Instagram [PITH_FULL_IMAGE:figures/full_fig_p014_12.png] view at source ↗

**Figure 13.** Figure 13: Confusion matrices comparing the computational classifications (predicted labels) with the human annota [PITH_FULL_IMAGE:figures/full_fig_p020_13.png] view at source ↗

**Figure 14.** Figure 14: Three confusion matrices comparing the person count performance across the four labels and between the [PITH_FULL_IMAGE:figures/full_fig_p020_14.png] view at source ↗

read the original abstract

This paper presents a computational case study that evaluates the capabilities of specialized machine learning models and emerging multimodal large language models for Visual Political Communication (VPC) analysis. Focusing on concentrated visibility in Instagram stories and posts during the 2021 German federal election campaign, we compare the performance of traditional computer vision models (FaceNet512, RetinaFace, Google Cloud Vision) with a multimodal large language model (GPT-4o) in identifying front-runner politicians and counting individuals in images. GPT-4o outperformed the other models, achieving a macro F1-score of 0.89 for face recognition and 0.86 for person counting in stories. These findings demonstrate the potential of advanced AI systems to scale and refine visual content analysis in political communication while highlighting methodological considerations for future research.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

GPT-4o beats the CV baselines on politician recognition and person counting in 2021 German election Instagram stories, but the F1 scores rest on unvalidated manual labels with no reported dataset size or agreement stats.

read the letter

The main thing to know is that this paper finds GPT-4o beating the traditional computer vision models at two tasks on 2021 German election Instagram content: recognizing specific politicians and counting the number of people in the images. It gives macro F1 scores of 0.89 and 0.86 respectively. The direct comparison using real campaign stories is what stands out as new. The paper does a decent job laying out the problem in visual political communication and showing how a multimodal LLM can be applied without much customization. It highlights the potential for these models to handle the informal, vertical format of stories better than older face detection tools like FaceNet or RetinaFace. Where it falls short is the lack of detail on the evaluation data. There is no mention of the total number of images, how the ground-truth labels for identities and counts were produced, if multiple people did the labeling, or any measure of consistency between labelers. Edge cases such as blurry images or multiple faces are not discussed in terms of errors. This makes it difficult to assess whether the reported advantage is robust or could be influenced by labeling choices. The concern about missing validation for the manual annotations is valid based on what is shown. Researchers working on large-scale analysis of social media in politics would find this relevant, as it offers a practical example of using current AI tools for content coding. I would send it out for peer review. The topic is timely and the approach is accessible, but the reviewers should ask for the dataset and annotation information to make the claims evaluable.

Referee Report

2 major / 2 minor

Summary. The paper presents an empirical case study comparing traditional computer vision models (FaceNet512, RetinaFace, Google Cloud Vision) against the multimodal LLM GPT-4o for two tasks in visual political communication: identifying front-runner politicians via face recognition and counting individuals in Instagram stories/posts from the 2021 German federal election campaign. It reports that GPT-4o achieves the highest performance with macro F1-scores of 0.89 (face recognition) and 0.86 (person counting in stories), arguing this demonstrates the potential of MLLMs to scale VPC analysis.

Significance. If the ground-truth labels prove reliable, the work provides a useful head-to-head comparison showing MLLMs can outperform specialized CV pipelines on real-world political imagery, with direct implications for scaling computational analysis in political communication and social media studies. The empirical focus and concrete F1 metrics are strengths, but the absence of dataset scale, annotation validation, and error analysis currently prevents the claims from being fully interpretable or generalizable.

major comments (2)

[Methods / Data and Annotation] The manuscript reports macro F1-scores of 0.89 and 0.86 for GPT-4o but supplies no dataset size, number of images/stories, annotation protocol, number of labelers, or inter-annotator agreement statistics for the politician identity and person-count ground truth. This information is load-bearing for the central performance claims and must be added to allow evaluation of whether labeling biases could favor one model.
[Results] No error analysis, confusion matrices, or discussion of edge cases (low-resolution stories, occluded faces, group shots, or ambiguous politician identities) is provided in the results. Without this, it is impossible to determine whether GPT-4o's reported advantage is robust or an artifact of the manual labeling process.

minor comments (2)

[Introduction] The abstract and introduction use 'concentrated visibility' without a concise definition or citation to the political communication literature; adding one sentence would improve accessibility.
[Results] Table or figure captions for the model comparison results should explicitly state the number of test instances and the exact definition of 'macro F1' used.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. The comments identify key gaps in transparency that we will address through revision to strengthen the interpretability of our results.

read point-by-point responses

Referee: [Methods / Data and Annotation] The manuscript reports macro F1-scores of 0.89 and 0.86 for GPT-4o but supplies no dataset size, number of images/stories, annotation protocol, number of labelers, or inter-annotator agreement statistics for the politician identity and person-count ground truth. This information is load-bearing for the central performance claims and must be added to allow evaluation of whether labeling biases could favor one model.

Authors: We agree that these methodological details are essential for assessing ground-truth reliability and potential biases. The revised manuscript will add a dedicated subsection in the Methods section that reports the total number of Instagram stories and posts analyzed, the data collection period and sampling strategy from the 2021 German federal election, the annotation protocol (including guidelines provided to labelers), the number of annotators, and inter-annotator agreement statistics for both politician identity and person-count labels. revision: yes
Referee: [Results] No error analysis, confusion matrices, or discussion of edge cases (low-resolution stories, occluded faces, group shots, or ambiguous politician identities) is provided in the results. Without this, it is impossible to determine whether GPT-4o's reported advantage is robust or an artifact of the manual labeling process.

Authors: We acknowledge the value of error analysis for demonstrating robustness. The revised Results section will include confusion matrices for both tasks and a new qualitative error analysis subsection that examines performance on the specified edge cases (low-resolution stories, occluded faces, group shots, and ambiguous identities), comparing GPT-4o against the baseline models and discussing any patterns that could relate to labeling artifacts. revision: yes

Circularity Check

0 steps flagged

No circularity: pure empirical model comparison on independent annotations

full rationale

The paper conducts an empirical evaluation of off-the-shelf and multimodal models (FaceNet512, RetinaFace, Google Cloud Vision, GPT-4o) against manually created ground-truth labels for politician identification and person counting on Instagram images. No equations, derivations, parameter fitting, or predictions are presented; performance is measured via standard F1 scores on held-out annotations. No self-citations serve as load-bearing premises for any claim, and no result reduces to its own inputs by construction. The analysis is self-contained against external benchmarks (model outputs vs. human labels).

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an empirical case study with no mathematical derivations, free parameters, or postulated entities.

pith-pipeline@v0.9.0 · 5434 in / 1048 out tokens · 42611 ms · 2026-05-10T02:32:25.640723+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

194 extracted references · 139 canonical work pages

[1]

doi:10.1145/3717867.3717881 , url =

Achmann-Denkler, Michael and Haim, Mario and Wolff, Christian , year = 2025, booktitle =. doi:10.1145/3717867.3717881 , url =

work page doi:10.1145/3717867.3717881 2025
[2]

doi:10.17645/pag.8727 , issn =

Bene, Márton and Magin, Melanie and Haßler, Jörg , year = 2024, month = jul, journal =. doi:10.17645/pag.8727 , issn =

work page doi:10.17645/pag.8727 2024
[3]

doi:10.5117/ccr2022.1.005.jurg , issn =

Jürgens, Pascal and Meltzer, Christine E and Scharkow, Michael , year = 2022, month = feb, journal =. doi:10.5117/ccr2022.1.005.jurg , issn =

work page doi:10.5117/ccr2022.1.005.jurg 2022
[4]

Yue, Xiang and Ni, Yuansheng and Zhang, Kai and Zheng, Tianyu and Liu, Ruoqi and Zhang, Ge and Stevens, Samuel and Jiang, Dongfu and Ren, Weiming and Sun, Yuxuan and Wei, Cong and Yu, Botao and Yuan, Ruibin and Sun, Renliang and Yin, Ming and Zheng, Boyuan and Yang, Zhenzhu and Liu, Yibo and Huang, Wenhao and Sun, Huan and Su, Yu and Chen, Wenhu , year = ...

2023
[5]

Veneti, Anastasia , year = 2023, month = jan, publisher =

2023
[6]

Peng, Yilang and Lu, Yingdan , year = 2023, booktitle =

2023
[8]

doi:10.4337/9781789906769.00023 , isbn =

Schwemmer, Carsten and Unger, Saïd and Heiberger, Raphael , year = 2023, month = mar, booktitle =. doi:10.4337/9781789906769.00023 , isbn =

work page doi:10.4337/9781789906769.00023 2023
[9]

doi:10.1109/34.895972 , issn =

Smeulders, A W M and Worring, M and Santini, S and Gupta, A and Jain, R , year = 2000, journal =. doi:10.1109/34.895972 , issn =

work page doi:10.1109/34.895972 2000
[10]

doi:10.1080/19331681.2021.1928579 , issn =

Schmøkel, Rasmus and Bossetta, Michael , year = 2022, month = jan, journal =. doi:10.1080/19331681.2021.1928579 , issn =

work page doi:10.1080/19331681.2021.1928579 2022
[11]

doi:10.4337/9781800376939.00021 , isbn =

Raynauld, Vincent and Lalancette, Mireille , year = 2023, month = jan, booktitle =. doi:10.4337/9781800376939.00021 , isbn =

work page doi:10.4337/9781800376939.00021 2023
[12]

Political Communication , publisher =

Bossetta, Michael and Schm. Political Communication , publisher =. doi:10.1080/10584609.2022.2128949 , issn =

work page doi:10.1080/10584609.2022.2128949 2022
[13]

doi:10.7551/mitpress/14046.001.0001 , isbn = 9780262546133, url =

Arnold, Taylor and Tilton, Lauren , year = 2023, month = oct, publisher =. doi:10.7551/mitpress/14046.001.0001 , isbn = 9780262546133, url =

work page doi:10.7551/mitpress/14046.001.0001 2023
[14]

Achmann-Denkler, Michael and Fehle, Jakob and Haim, Mario and Wolff, Christian , year = 2024, journal =

2024
[15]

Goffman, Erving , year = 1956, publisher =

1956
[16]

doi:10.4337/9781800376939.00020 , isbn =

Šimunjak, Maja , year = 2023, month = jan, booktitle =. doi:10.4337/9781800376939.00020 , isbn =

work page doi:10.4337/9781800376939.00020 2023
[17]

Marcus, G E and Neuman, W R and Mackuen, M , year = 2000, publisher =

2000
[18]

doi:10.1080/1369118x.2021.1962942 , issn =

Brands, Charlotte and Kruikemeier, Sanne and Trilling, Damian , year = 2021, month = oct, journal =. doi:10.1080/1369118x.2021.1962942 , issn =

work page doi:10.1080/1369118x.2021.1962942 2021
[19]

doi:10.1177/1940161220937239 , issn =

Lindholm, Jenny and Carlson, Tom and Högväg, Joachim , year = 2021, month = jan, journal =. doi:10.1177/1940161220937239 , issn =

work page doi:10.1177/1940161220937239 2021
[20]

doi:10.1177/2378023120967171 , issn =

Schwemmer, Carsten and Knight, Carly and Bello-Pardo, Emily D and Oklobdzija, Stan and Schoonvelde, Martijn and Lockhart, Jeffrey W , year = 2020, month = nov, journal =. doi:10.1177/2378023120967171 , issn =

work page doi:10.1177/2378023120967171 2020
[21]

doi:10.1080/1369118X.2022.2027500 , issn =

Parmelee, John H and Perkins, Stephynie C and Beasley, Berrin , year = 2023, month = jul, journal =. doi:10.1080/1369118X.2022.2027500 , issn =

work page doi:10.1080/1369118x.2022.2027500 2023
[22]

doi:10.1017/pan.2022.26 , issn =

Tarr, Alexander and Hwang, June and Imai, Kosuke , year = 2022, month = nov, journal =. doi:10.1017/pan.2022.26 , issn =

work page doi:10.1017/pan.2022.26 2022
[23]

doi:10.1609/icwsm.v14i1.7338 , issn =

Xi, Nan and Ma, Di and Liou, Marcus and Steinert-Threlkeld, Zachary C and Anastasopoulos, Jason and Joo, Jungseock , year = 2020, month = may, journal =. doi:10.1609/icwsm.v14i1.7338 , issn =

work page doi:10.1609/icwsm.v14i1.7338 2020
[24]

doi:10.1177/14614448211038761 , issn =

Chen, Yan and Sherren, Kate and Smit, Michael and Lee, Kyung Young , year = 2023, month = apr, journal =. doi:10.1177/14614448211038761 , issn =

work page doi:10.1177/14614448211038761 2023
[25]

doi:10.5617/dhnbpub.10647 , issn =

Achmann, Michael and Wolff, Christian , year = 2023, month = oct, journal =. doi:10.5617/dhnbpub.10647 , issn =

work page doi:10.5617/dhnbpub.10647 2023
[26]

Zhan, Q., Fang, R., Panchal, H

Ziems, Caleb and Held, William and Shaikh, Omar and Chen, Jiaao and Zhang, Zhehao and Yang, Diyi , year = 2024, month = mar, journal =. doi:10.1162/coli\_a\_00502 , issn =

work page doi:10.1162/coli 2024
[27]

doi:10.1017/pan.2023.33 , issn =

Girbau, Andreu and Kobayashi, Tetsuro and Renoust, Benjamin and Matsui, Yusuke and Satoh, Shin'ichi , year = 2024, month = apr, journal =. doi:10.1017/pan.2023.33 , issn =

work page doi:10.1017/pan.2023.33 2024
[28]

doi:10.2478/njms-2024-0008 , issn =

Magin, Melanie and Haßler, Jörg and Larsson, Anders Olof and Skogerbø, Eli , year = 2024, month = aug, journal =. doi:10.2478/njms-2024-0008 , issn =

work page doi:10.2478/njms-2024-0008 2024
[29]

doi:10.1016/j.linged.2019.05.003 , issn =

Georgakopoulou, Alex , year = 2021, month = apr, journal =. doi:10.1016/j.linged.2019.05.003 , issn =

work page doi:10.1016/j.linged.2019.05.003 2021
[30]

Joo, Jungseock and Bucy, E and Seidel, Claudia , year = 2019, month = sep, journal =

2019
[31]

doi:10.1111/comt.12048 , issn =

Geise, Stephanie and Baden, Christian , year = 2015, month = feb, journal =. doi:10.1111/comt.12048 , issn =

work page doi:10.1111/comt.12048 2015
[32]

Politische Werbung und koordiniertes Verhalten in Sozialen Medien im Vorfeld der Bundestagswahl 2021 , author =

2021
[33]

doi:10.1177/08944393241227554 , issn =

Towner, Terri L and Muñoz, Caroline L , year = 2024, month = feb, journal =. doi:10.1177/08944393241227554 , issn =

work page doi:10.1177/08944393241227554 2024
[34]

doi:10.1007/s10919-009-0082-1 , issn =

Olivola, Christopher Y and Todorov, Alexander , year = 2010, month = jun, journal =. doi:10.1007/s10919-009-0082-1 , issn =

work page doi:10.1007/s10919-009-0082-1 2010
[35]

doi:10.15581/003.33.37241 , issn =

Sampietro, Agnese and Sánchez-Castillo, Sebastián , year = 2020, month = jan, journal =. doi:10.15581/003.33.37241 , issn =

work page doi:10.15581/003.33.37241 2020
[36]

doi:10.1177/1461444816686103 , issn =

McGregor, Shannon C , year = 2018, month = mar, journal =. doi:10.1177/1461444816686103 , issn =

work page doi:10.1177/1461444816686103 2018
[37]

doi:10.1080/1369118x.2016.1167228 , issn =

McGregor, Shannon C and Lawrence, Regina G and Cardona, Arielle , year = 2017, month = feb, journal =. doi:10.1080/1369118x.2016.1167228 , issn =

work page doi:10.1080/1369118x.2016.1167228 2017
[38]

doi:10.1177/2056305116662179 , issn =

Filimonov, Kirill and Russmann, Uta and Svensson, Jakob , year = 2016, month = jul, journal =. doi:10.1177/2056305116662179 , issn =

work page doi:10.1177/2056305116662179 2016
[39]

doi:10.4337/9781800376939.00016 , isbn =

Steffan, Dennis , year = 2023, month = jan, booktitle =. doi:10.4337/9781800376939.00016 , isbn =

work page doi:10.4337/9781800376939.00016 2023
[40]

Steffan, Dennis , year = 2020, month = may, journal =

2020
[41]

doi:10.5771/9783748925415 , isbn = 9783848781249, abstract =

Steffan, Dennis , year = 2021, month = jul, publisher =. doi:10.5771/9783748925415 , isbn = 9783848781249, abstract =

work page doi:10.5771/9783748925415 2021
[43]

doi:10.1177/1461444812457333 , issn =

Hermans, Liesbeth and Vergeer, Maurice , year = 2013, month = feb, journal =. doi:10.1177/1461444812457333 , issn =

work page doi:10.1177/1461444812457333 2013
[44]

doi:10.1177/1065912918786805 , issn =

Casas, Andreu and Williams, Nora Webb , year = 2019, month = jun, journal =. doi:10.1177/1065912918786805 , issn =

work page doi:10.1177/1065912918786805 2019
[45]

doi:10.1109/iccv.2015.423 , isbn = 9781467383912, url =

Joo, Jungseock and Steen, Francis F and Zhu, Song-Chun , year = 2015, month = dec, booktitle =. doi:10.1109/iccv.2015.423 , isbn = 9781467383912, url =

work page doi:10.1109/iccv.2015.423 2015
[46]

doi:10.1017/s0022381610000939 , issn =

Valentino, Nicholas A and Brader, Ted and Groenendyk, Eric W and Gregorowicz, Krysha and Hutchings, Vincent L , year = 2011, month = jan, journal =. doi:10.1017/s0022381610000939 , issn =

work page doi:10.1017/s0022381610000939 2011
[47]

doi:10.1080/15205436.2010.530385 , issn =

Tukachinsky, Riva and Mastro, Dana and King, Aimee , year = 2011, month = nov, journal =. doi:10.1080/15205436.2010.530385 , issn =

work page doi:10.1080/15205436.2010.530385 2011
[48]

doi:10.1177/0002716296546001008 , issn =

Graber, Doris A , year = 1996, month = jul, journal =. doi:10.1177/0002716296546001008 , issn =

work page doi:10.1177/0002716296546001008 1996
[49]

doi:10.5117/ccr2022.1.001.joo , issn =

Joo, Jungseock and Steinert-Threlkeld, Zachary C , year = 2022, month = feb, journal =. doi:10.5117/ccr2022.1.001.joo , issn =

work page doi:10.5117/ccr2022.1.001.joo 2022
[50]

doi:10.1080/19312458.2023.2285765 , issn =

Peng, Yilang and Lock, Irina and Ali Salah, Albert , year = 2024, month = apr, journal =. doi:10.1080/19312458.2023.2277956 , issn =

work page doi:10.1080/19312458.2023.2277956 2024
[51]

Poguntke, Thomas and Webb, Paul D , year = 2005, publisher =

2005
[52]

doi:10.1080/23808985.2010.11679101 , issn =

Adam, Silke and Maier, Michaela , year = 2010, month = jan, journal =. doi:10.1080/23808985.2010.11679101 , issn =

work page doi:10.1080/23808985.2010.11679101 2010
[53]

doi:10.1007/978-3-030-18729-3 , isbn =

work page doi:10.1007/978-3-030-18729-3
[54]

Python 3 Reference Manual , author =
[55]

doi:10.1386/jvpc\_00027\_1 , issn =

Farkas, Xénia , year = 2023, month = oct, journal =. doi:10.1386/jvpc\_00027\_1 , issn =

work page doi:10.1386/jvpc 2023
[56]

doi:10.1177/14614448231224031 , issn =

Geise, Stephanie and Maubach, Katharina and Boettcher Eli, Alena , year = 2024, month = mar, journal =. doi:10.1177/14614448231224031 , issn =

work page doi:10.1177/14614448231224031 2024
[57]

A benchmark of facial recognition pipelines and co-usability performances of modules.Journal of Information Technologies, 17(2):95–107, 2024

A Benchmark of Facial Recognition Pipelines and Co-Usability Performances of Modules , author =. Bilisim Teknolojileri Dergisi , publisher =. doi:10.17671/gazibtd.1399077 , url =

work page doi:10.17671/gazibtd.1399077
[58]

Pryzant, Reid and Iter, Dan and Li, Jerry and Lee, Yin Tat and Zhu, Chenguang and Zeng, Michael , year = 2023, month = may, url =

2023
[59]

doi:10.1080/15358593.2011.653504 , url =

Schill, Dan , year = 2012, month = apr, journal =. doi:10.1080/15358593.2011.653504 , url =

work page doi:10.1080/15358593.2011.653504 2012
[60]

doi:10.1080/01972243.1994.9960162 , issn =

Agre, Philip E , year = 1994, month = apr, journal =. doi:10.1080/01972243.1994.9960162 , issn =

work page doi:10.1080/01972243.1994.9960162 1994
[61]

doi:10.1007/BF00988269 , issn =

Altheide, David L , year = 1987, month = mar, journal =. doi:10.1007/BF00988269 , issn =

work page doi:10.1007/bf00988269 1987
[62]

Amancio, Marina , year = 2017, url =

2017
[63]

doi:10.1080/19312458.2020.1810648 , issn =

Araujo, Theo and Lock, Irina and van de Velde, Bob , year = 2020, month = oct, journal =. doi:10.1080/19312458.2020.1810648 , issn =

work page doi:10.1080/19312458.2020.1810648 2020
[64]

doi:10.1093/llc/fqz013 , issn =

Arnold, Taylor and Tilton, Lauren , year = 2019, month = dec, journal =. doi:10.1093/llc/fqz013 , issn =

work page doi:10.1093/llc/fqz013 2019
[65]

doi:10.1080/19312458.2021.2015574 , issn =

Baden, Christian and Pipal, Christian and Schoonvelde, Martijn and van der Velden, Mariken A C G , year = 2022, month = jan, journal =. doi:10.1080/19312458.2021.2015574 , issn =

work page doi:10.1080/19312458.2021.2015574 2022
[66]

doi:10.1177/1461444820960071 , issn =

Bainotti, Lucia and Caliandro, Alessandro and Gandini, Alessandro , year = 2020, month = sep, journal =. doi:10.1177/1461444820960071 , issn =

work page doi:10.1177/1461444820960071 2020
[67]

doi:10.1080/15377857.2021.1892901 , issn =

Bast, Jennifer , year = 2021, month = mar, journal =. doi:10.1080/15377857.2021.1892901 , issn =

work page doi:10.1080/15377857.2021.1892901 2021
[68]

Bast, Jennifer , year = 2021, month = jul, journal =

2021
[69]

doi:10.1080/1369118X.2015.1084349 , issn =

Bayer, Joseph B and Ellison, Nicole B and Schoenebeck, Sarita Y and Falk, Emily B , year = 2016, month = jul, journal =. doi:10.1080/1369118X.2015.1084349 , issn =

work page doi:10.1080/1369118x.2015.1084349 2016
[70]

doi:10.1177/1077699018763307 , issn =

Bossetta, Michael , year = 2018, month = jun, journal =. doi:10.1177/1077699018763307 , issn =

work page doi:10.1177/1077699018763307 2018
[71]

doi:10.1177/14614448211009504 , issn =

Boulianne, Shelley and Larsson, Anders Olof , year = 2021, month = apr, journal =. doi:10.1177/14614448211009504 , issn =

work page doi:10.1177/14614448211009504 2021
[72]

doi:10.1177/0170840618772611 , issn =

Boxenbaum, Eva and Jones, Candace and Meyer, Renate E and Svejenova, Silviya , year = 2018, month = jun, journal =. doi:10.1177/0170840618772611 , issn =

work page doi:10.1177/0170840618772611 2018
[73]

Profesional de la informaci

Coromina,. Profesional de la informaci. doi:10.3145/epi.2018.sep.05 , issn =

work page doi:10.3145/epi.2018.sep.05 2018
[74]

doi:10.1007/978-3-319-45270-8\_8 , isbn =

Ekman, Mattias and Widholm, Andreas , year = 2017, booktitle =. doi:10.1007/978-3-319-45270-8\_8 , isbn =

work page doi:10.1007/978-3-319-45270-8 2017
[75]

doi:10.1177/0267323113516727 , issn =

Holtz-Bacha, Christina and Langer, Ana Ines and Merkle, Susanne , year = 2014, month = apr, journal =. doi:10.1177/0267323113516727 , issn =

work page doi:10.1177/0267323113516727 2014
[76]

doi:10.1177/1354856517741132 , issn =

Larsson, Anders Olof , year = 2019, month = dec, journal =. doi:10.1177/1354856517741132 , issn =

work page doi:10.1177/1354856517741132 2019
[77]

doi:10.1177/1940161220964769 , issn =

Peng, Yilang , year = 2021, month = jan, journal =. doi:10.1177/1940161220964769 , issn =

work page doi:10.1177/1940161220964769 2021
[78]

doi:10.1386/nl.15.1.15\_1 , issn =

Ekman, Mattias and Widholm, Andreas , year = 2017, journal =. doi:10.1386/nl.15.1.15\_1 , issn =

work page doi:10.1386/nl.15.1.15 2017
[79]

doi:10.1080/1369118x.2019.1581244 , issn =

Metz, Manon and Kruikemeier, Sanne and Lecheler, Sophie , year = 2020, month = aug, journal =. doi:10.1080/1369118x.2019.1581244 , issn =

work page doi:10.1080/1369118x.2019.1581244 2020
[80]

moco , url=

Deng, Jiankang and Guo, Jia and Ververas, Evangelos and Kotsia, Irene and Zafeiriou, Stefanos , year = 2020, month = jun, booktitle =. doi:10.1109/cvpr42600.2020.00525 , isbn = 9781728171685, url =

work page doi:10.1109/cvpr42600.2020.00525 2020
[81]

doi:10.1080/1369118X.2016.1198411 , issn =

Bene, Marton , year = 2017, month = apr, journal =. doi:10.1080/1369118X.2016.1198411 , issn =

work page doi:10.1080/1369118x.2016.1198411 2017
[82]

Huang, Gary B and Mattar, Marwan and Berg, Tamara and Learned-Miller, Eric , year = 2008, month = oct, booktitle =

2008

Showing first 80 references.