In 2010, a paper entitled "From Obscurity to Prominence in Minutes: Political Speech and Real-time search" won the Best Paper Prize of the Web Science 2010 Conference. Among its findings were the discovery and documentation of what was termed a "Twitter-bomb", an organized effort to spread misinformation about the democratic candidate Martha Coakley through anonymous Twitter accounts. In this paper, after summarizing the details of that event, we outline the recipe of how social networks are used to spread misinformation. One of the most important steps in such a recipe is the "infiltration" of a community of users who are already engaged in conversations about a topic, to use them as organic spreaders of misinformation in their extended subnetworks. Then, we take this misinformation spreading recipe and indicate how it was successfully used to spread fake news during the 2016 U.S. Presidential Election. The main differences between the scenarios are the use of Facebook instead of Twitter, and the respective motivations (in 2010: political influence; in 2016: financial benefit through online advertising). After situating these events in the broader context of exploiting the Web, we seize this opportunity to address limitations of the reach of research findings and to start a conversation about how communities of researchers can increase their impact on real-world societal issues.
Abstract Online trolling, disinformation, and deception are posing an existential threat to democracy. Informed by the online disinhibition theory and research on the ideological asymmetry between Democrats and Republicans, we examined how the extent and style of trolling varies across social media platforms, by analyzing comments on posts by two media channels (CNN and Fox News) on three social media platforms (Facebook, Instagram, and Twitter). We found differences in the style and extent of trolling across platforms and between media channels, with more trolling on articles posted by Fox News than by CNN, and a different trolling style on Twitter than Facebook or Instagram. Our study demonstrates a delicate balance between the socio‐technical factors that are enabling and hindering trolling. While some platforms and government agencies believe in removing anonymity to regulate online harm, this paper makes a significant contribution against that view.
暂无摘要（点击查看原文获取完整内容）
ABSTRACT Recent studies propose that limited investor attention causes market underreactions. This paper directly tests this explanation by measuring the information load faced by investors. The investor distraction hypothesis holds that extraneous news inhibits market reactions to relevant news. We find that the immediate price and volume reaction to a firm's earnings surprise is much weaker, and post‐announcement drift much stronger, when a greater number of same‐day earnings announcements are made by other firms. We evaluate the economic importance of distraction effects through a trading strategy, which yields substantial alphas. Industry‐unrelated news and large earnings surprises have a stronger distracting effect.
Sébastien Jean, Kyunghyun Cho, Roland Memisevic, Yoshua Bengio. Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2015.
This chapter reviews and synthesizes the literature pertaining to the supply of disinformation, misinformation, and fake news in three main sections. First, we define and draw boundaries between the concepts of disinformation, misinformation, and fake news. Next, we offer an extensive discussion of the supply of these concepts. Specifically, we discuss how different actor groups (i.e. political actors, media actors, and citizens) create and disseminate inaccurate information. Here, we focus on these actors’ motives and how the impact of falsehoods might differ when it is supplied by political actors, media actors, or citizens. In the third section, we explain how the supply of actual disinformation, misinformation, and fake news is accompanied by a second mechanism, i.e. the perceived supply of falsehood, referring to growing beliefs that perfectly factual information is incorrect. We connect this to the ubiquitous debate about the prevalence of mis- and disinformation in many democracies, and specifically, to the weaponization of the term “fake news” by politicians to systematically discredit public trust in news media. We conclude by emphasizing the difficulties research faces in investigating the supply of disinformation, misinformation, and fake news and suggesting several avenues for future research on this crucial matter.
Harmful lies are nothing new. But the ability to distort reality has taken an exponential leap forward with “deep fake” technology. This capability makes it possible to create audio and video of real people saying and doing things they never said or did. Machine learning techniques are escalating the technology’s sophistication, making deep fakes ever more realistic and increasingly resistant to detection. Deep-fake technology has characteristics that enable rapid and widespread diffusion, putting it into the hands of both sophisticated and unsophisticated actors. While deep-fake technology will bring with it certain benefits, it also will introduce many harms. The marketplace of ideas already suffers from truth decay as our networked information environment interacts in toxic ways with our cognitive biases. Deep fakes will exacerbate this problem significantly. Individuals and businesses will face novel forms of exploitation, intimidation, and personal sabotage. The risks to our democracy and to national security are profound as well. Our aim is to provide the first in-depth assessment of the causes and consequences of this disruptive technological change, and to explore the existing and potential tools for responding to it. We survey a broad array of responses, including: the role of technological solutions; criminal penalties, civil liability, and regulatory action; military and covert-action responses; economic sanctions; and market developments. We cover the waterfront from immunities to immutable authentication trails, offering recommendations to improve law and policy and anticipating the pitfalls embedded in various solutions.
Abstract The rise of “fake news” is a major concern in contemporary Western democracies. Yet, research on the psychological motivations behind the spread of political fake news on social media is surprisingly limited. Are citizens who share fake news ignorant and lazy? Are they fueled by sinister motives, seeking to disrupt the social status quo? Or do they seek to attack partisan opponents in an increasingly polarized political environment? This article is the first to test these competing hypotheses based on a careful mapping of psychological profiles of over 2,300 American Twitter users linked to behavioral sharing data and sentiment analyses of more than 500,000 news story headlines. The findings contradict the ignorance perspective but provide some support for the disruption perspective and strong support for the partisan polarization perspective. Thus, individuals who report hating their political opponents are the most likely to share political fake news and selectively share content that is useful for derogating these opponents. Overall, our findings show that fake news sharing is fueled by the same psychological motivations that drive other forms of partisan behavior, including sharing partisan news from traditional and credible news sources.
Topological data analysis has recently found applications in various areas of science, such as computer vision and understanding of protein folding. However, applications of topological data analysis to natural language processing remain under-researched. This study applies topological data analysis to a particular natural language processing task: fake news detection. We have found that deep learning models are more accurate in this task than topological data analysis. However, assembling a deep learning model with topological data analysis significantly improves the model’s accuracy if the available training set is very small.
Automatically generating coherent and semantically meaningful text has many applications in machine translation, dialogue systems, image captioning, etc. Recently, by combining with policy gradient, Generative Adversarial Nets(GAN) that use a discriminative model to guide the training of the generative model as a reinforcement learning policy has shown promising results in text generation. However, the scalar guiding signal is only available after the entire text has been generated and lacks intermediate information about text structure during the generative process. As such, it limits its success when the length of the generated text samples is long (more than 20 words). In this paper, we propose a new framework, called LeakGAN, to address the problem for long text generation. We allow the discriminative net to leak its own high-level extracted features to the generative net to further help the guidance. The generator incorporates such informative signals into all generation steps through an additional MANAGER module, which takes the extracted features of current generated words and outputs a latent vector to guide the WORKER module for next-word generation.Our extensive experiments on synthetic data and various real-world tasks with Turing test demonstrate that LeakGAN is highly effective in long text generation and also improves the performance in short text generation scenarios. More importantly, without any supervision, LeakGAN would be able to implicitly learn sentence structures only through the interaction between MANAGER and WORKER.
The purpose of business sentiment analysis is to determine the emotions or attitudes expressed toward the company, products, services, personnel, or events. Text analysis are the simplest and most developed types of sentiment analysis so far. The text-based business sentiment analysis still has some unresolved challenges. For example, the machine learning algorithms are unable to recognize double meanings, jokes and allusions. The regional differences between language and non-native speech structures cannot be explained. To solve this problem, an undirected weighted graph is constructed for news topics. The sentences in an article are modeled as nodes, and the normalized sentence similarity is used as the link of the nodes, which can help avoid the influence of sentence length on the summary results. In the topic extraction process, the keywords are not limited to a single word, to achieve the purpose of improving the readability of the abstract. To improve the accuracy of sentiment classification, this work proposes a robust news mining-based business sentiment analysis framework, called BuSeD. It contains two main stages: (1) news collection and preprocessing, and (2) feature extraction and sentiment classification. In the first stage, the news is collected by using crawler tools. The news dataset is then preprocessed by reducing noises. In the second stage, topics in each article is extracted by using traditional topic extraction tools. And then a convolutional neural network (CNN)-based text analyzing model is designed to analyze news from sentence level. We conduct comprehensive experiments to evaluate the performance of BuSeD for sentiment classification. Compared with four classical classification algorithms, the proposed CNN-based classification model of BuSeD achieves the highest F1 scores. We also present a quantitative trading application based on sentiment analysis to validate BuSeD, which indicates that the news-based business sentiment analysis has high economic application value.
We propose a formal test of the hypothesis that energy prices are predetermined with respect to U.S. macroeconomic aggregates. The test is based on regressing changes in daily energy prices on daily news from U.S. macroeconomic data releases. Using a wide range of macroeconomic news, we find no compelling evidence of feedback at daily or monthly horizons, contradicting the view that energy prices respond instantaneously to macroeconomic news and consistent with the commonly used identifying assumption that there is no feedback from U.S. macroeconomic aggregates to monthly innovations in energy prices.
We characterize the response of U.S., German and British stock, bond and foreign exchange markets to real-time U.S. macroeconomic news. Our analysis is based on a unique data set of high-frequency futures returns for each of the markets. We find that news surprises produce conditional mean jumps; hence high-frequency stock, bond and exchange rate dynamics are linked to fundamentals. The details of the linkages are particularly intriguing as regards equity markets. We show that equity markets react differently to the same news depending on the state of the U.S. economy, with bad news having a positive impact during expansions and the traditionally-expected negative impact during recessions. We rationalize this by temporal variation in the competing "cash flow" and "discount rate" effects for equity valuation. This finding also helps explain the apparent time-varying correlation between stock and bond returns, and the relatively small equity market news announcement effect when averaged across expansions and recessions. Hence, while our results confirm previous unconditional rankings suggesting that bond markets almost uniformly react most strongly to macroeconomic news, followed by foreign exchange and then equity markets, importantly when conditioning on the state of the economy the foreign exchange and equity markets appear equally responsive. Lastly, relying on the pronounced heteroskedasticity in the new high-frequency data, we also document important contemporaneous linkages across all markets and countries overand-above the direct news announcement effects.
暂无摘要（点击查看原文获取完整内容）
暂无摘要（点击查看原文获取完整内容）
Abstract We investigate the relationship between investor attention and financial market anomalies. We find that anomaly returns tend to be higher following high-attention days. The result is robust after controlling for the effect of news and in a natural experiment setting in which a stock market regulation and rounding errors generate exogenous variations in attention. An analysis of order imbalances suggests that large traders trade on anomaly signals more aggressively upon observing higher attention. We discuss the extent to which the findings are driven by inattention-driven underreaction, bias amplification, or coordinated arbitrage mechanisms, thereby providing insight into the understanding of anomalies.
This report reveals new insights about digital news consumption. Based on a representative survey of online news consumers across twelve countries – UK, US, Germany, France, Denmark, Spain, Italy, Finland, Ireland, Australia, Urban Brazil and Japan – the report tracks the changing digital news behaviour of consumers.
the literature, and discuss substantive lessons.
Abstract We examine empirically the response of bond returns and their volatility to good and bad macroeconomic news during expansions and recessions. We find that macroeconomic announcements are most important when they contain bad news for bond returns in expansions and, to a lesser extent, good news in contractions. In expansions, the bond market responds most strongly to bad news in non-farm payrolls, while in recessions good news about inflation is relatively more important. We also document that macroeconomic news impacts the volatility of bond returns at all maturities by increasing jump intensities and altering the jump size distribution.
Using a new dataset consisting of six years of real-time exchange rate quotations, macroeconomic expectations, and macroeconomic realizations (announcements), we characterize the conditional means of U.S. dollar spot exchange rates versus German Mark, British Pound, Japanese Yen, Swiss Franc, and the Euro. In particular, we find that announcement surprises (that is, divergences between expectations and realizations, or "news") produce conditional mean jumps; hence high-frequency exchange rate dynamics are linked to fundamentals. The details of the linkage are intriguing and include announcement timing and sign effects. The sign effect refers to the fact that the market reacts to news in an asymmetric fashion: bad news has greater impact than good news, which we relate to recent theoretical work on information processing and price discovery.