Peer-reviewed journal articles

Hilbert, M., Thakur, A., Flores, P. M., Zhang, X., Bhan, J. Y., Bernhard, P., & Ji, F. (2024). 8–10% of algorithmic recommendations are ‘bad’, but… an exploratory risk-utility meta-analysis and its regulatory implications. International Journal of Information Management75, 102743.

We conducted a quantitatively coarse-grained, but wide-ranging evaluation of the frequency recommender algorithms provide ‘good’ and ‘bad’ recommendations, with a focus on the latter. We found 151 algorithmic audits from 33 studies that report fitting risk-utility statistics from YouTube, Google Search, Twitter, Facebook, TikTok, Amazon, and others. Our findings indicate that roughly 8–10% of algorithmic recommendations are ‘bad’, while about a quarter actively protect users from self-induced harm (‘do good’). This average is remarkably consistent across the audits, irrespective of the platform nor on the kind of risk (bias/ discrimination, mental health and child harm, misinformation, or political extremism). Algorithmic audits find negative feedback loops that can ensnare users into spirals of ‘bad’ recommendations (or being ‘dragged down the rabbit hole’), but also highlight an even larger likelihood of positive spirals of ‘good recommendations’. While our analysis refrains from any judgment of the causal consequences and severity of risks, the detected levels surpass those associated with many other consumer products. They are comparable to the risk levels of generic food defects monitored by public authorities such as the FDA or FSIS in the United States. Consequently, our findings inform the ongoing discussion regarding regulatory oversight of the potential risks posed by recommender algorithms.

Flores, P. M. & Hilbert, M. (2023). Lean-back and lean-forward online behaviors: The role of emotions in passive versus proactive information diffusion of social media content. Computers in Human Behavior, 147, 107841.

One of the main drivers of social media’s influence is the easy way of generating information cascades via forwarding messages. Two basic modes of forwarding information consist in sharing it without changes (lean-back) or adding or modifying the content of the original message (lean-forward). In this work, we study these two modes of online information sharing. Using data from six cases extracted from Twitter, in which retweets make up most of the content, we analyzed emotions in passive versus proactive information diffusion. Our findings show that emotional valence does not indicate significant differences between lean behaviors, while the activation level of emotions presents contrasts. For example, disgust is more intense in lean-back and anger in lean-forward behavior. We also find that in proactive lean-forward communication, disgust and joy synchronize with the emotions of lean-back messages. Finally, a causal analysis of lean-forward information sharing reveals that disgust provokes a consistent increment of itself while it also decreases the appearance of anger in political topics. We discuss the implications of our findings for the study of emotions in active versus passive forwarding of information in social media.

Flores, P. M. & Hilbert, M. (2023). Temporal communication dynamics in the aftermath of large-scale upheavals: do digital footprints reveal a stage model? Journal of Computational Social Science

It has long been theorized that the exchange of information in the aftermath of large-scale upheavals ensues dynamics that follow a stage model, which would be a societal equivalent of individuals’ psychological processing of traumatic events. Nowadays, a relevant portion of this informational exchange occurs on social media platforms. In this study, we use the digital footprint of three independent earthquakes to analyze their communication dynamics. We find empirical evidence of a stage model previously proposed by Pennebaker (Pennebaker in Handbook of mental control, Prentice-Hall Inc., Hoboken, 1993) in the aftermath of the earthquakes. In addition, we further explore the role of emotions within the model stages through time using natural language processing tools. Our results show that emotions with low activation levels, such as interest and sadness, are expressed in higher proportions and are the most useful for predicting the expression of emotions with higher activation levels. Employing newly available computational methods like digital trace data, natural language processing, clustering, and causal analysis, this study extends Pennebaker’s model from offline to online social communication.

Mujica, C., Grassau, D., Bachmann, I., Herrada, N., Flores, P. M. & Puente, M. S. (2020). Percepciones de la audiencia respecto del uso del melodrama en noticias por televisión: entre el entusiasmo y el desprecio. Palabra Clave, 23(4), e2341.

In Latin America, news programs frequently leverage melodramatic resources: targeting of personal stories, exacerbation of emotions, use of rhetorical figures, close-ups, dramatic music, among others. This phenomenon tends to be criticized by experts, professionals, and members of the audience who associate melodrama with sensationalism or the coverage of irrelevant events. This article presents the results of two studies carried out in Chile using qualitative methods (focus groups, in the first case, and in-depth interviews, in the second). They aimed to explore the audience’s perceptions of the use of melodramatic resources in the Chilean open television news and the characteristics they value and reject, and find differences in such perception regarding informative coverage of highly emotional news, such as disasters. The results show mixed perceptions, between criticism and optimism, as the participants in both studies highlight that, while excess melodrama bothers them, its inclusion in news programs develops empathy and moves them to action. As to the coverage of disasters, stories are perceived differently between those affected and not affected by their consequences.

Puente, S., Marín, H., Alvarez, P., Flores, P. M. & Grassau, D. (2019) Mental Health and Media Links Based on Hobfoll’s Five Elements to Overcome Trauma: The Case of the Chilean 2010 Earthquake. Disasters, 43 (3), 555-574.

This paper reviews the role of news with respect to the mental health of a population exposed to a disaster. It is based on the five essential elements of psychosocial care presented by Stevan E. Hobfoll et al. (2007) that can be introduced after a potentially traumatic event: promoting a sense of safety, calming, self and collective efficacy, connectedness, and hope. This study developed a method to relate these elements to television coverage and applied it to the stories (n=1,169) aired by the main networks in Chile in the 72 hours after an 8.8 magnitude earthquake struck on 27 February 2010. Of the five elements, promoting a sense of safety occurred most often (82.72 per cent), whereas the others were barely present (less than 10 per cent). The study argues that these elements can increase the possibility of framing the news, given that the audience watching can also be affected by a disaster.

Valenzuela, S., Puente, S. & Flores, P. M. (2017). Comparing Disaster News on Twitter and Television: An intermedia agenda setting perspective. Journal of Broadcasting & Electronic Media, 61 (4), 615-637.

Intermedia agenda setting predicts a high degree of convergence between news media agendas. However, the rise of social media forces a re-examination of this expectation. Using the 8.8-earthquake of February 27, 2010 in Chile as a case study, this article compares which topics were covered by professional journalists on broadcast news and Twitter, analyzing both cross-sectional and longitudinal trends. A positive, reinforcing influence was found among the journalistic agendas of TV and Twitter. However, counter to the idea that social media are echo chambers of traditional media, it was found that Twitter influences TV news more so than the other way around. Thus, the study provides an early lens into the agenda setting function of social media among television news professionals, and its findings are consistent with Twitter succeeding among journalists due to its provision of valuable information.

Book Chapters

Fernández, F. & Flores, P. M. (2015). En búsqueda del usuario final y propuesta visual de la plataforma (In search of a visual proposal for the final user). In S. Pellegrini (Ed), Ordenando el caos: Gestión y modelamiento de los procesos de la industria de la comunicación, (pp. 207-243), Santiago, Chile, Ediciones UC.

Manuscripts in Progres

Flores P. M. & Hilbert M. (Data collection) Examining the Impact of Affective States and Emotional Responses on Misinformation Video Spreading.

Flores, P. M., Jin, X. & Hilbert, M. (Data collection) Algorithmic news curation: A comparison between news media websites and social media platforms.

Flores P. M. (Design in progress) A simulation approach to emotional cascades and their impacts on reshaping social networks