The irresistible appeal of expressing extreme emotions may have first appeared in traditional media. It is well known that health research claims are exaggerated in press releases and news articles, with over 50% of certain university press releases exaggerated, and some news agencies 60% is exaggerated.Five. This is “spin,” defined as a particular reporting strategy that can emphasize, whether intentionally or unintentionally, the beneficial effects of an experimental treatment.6. “Sensationalism” is a close companion when reporting on popular topics. It is a discourse strategy that “packages” information into news headlines so that the news items are presented as more interesting, special, and relevant.7. As the competition for online customers intensified, these practices became more prevalent and refined into new genres such as “clickbait.” This is nothing but an amplified headline designed to get readers to click on a link.8, the hype in online reviews. Exaggerated reviews are always absolutely positive or negative.9.
As traditional media has given way to social media and everyone is a “media outlet” these days, the temptation to engage in attention-seeking behavior is felt more personally. Influencers are using beauty filters to make the products they’re promoting look more effective, and have been warned by the Advertising Standards Authority (ASA) as a result.Ten. Young people aged 11-18 were observed to exaggerate their behavior in an attempt to meet amplified claims about popularity11. Even the accidental use of certain words can lead readers to believe in causal relationships that may not exist.12. This is also known as emotional polarity and is a key feature of fake news. To make news persuasive, authors often express strong positive or negative emotions in their content.13,14. As a result, bizarre conspiracy theories that may once have been the preserve of a small minority are now routinely accorded the same credibility as evidence-backed science by large swaths of the population.15.
It is impossible to conduct experiments on real human populations to understand the causes of extreme polarization. Computational models provide an important tool to overcome this empirical limitation. Computational models have been used for decades to study opinion dynamics, and early studies often focused on consensus building.16, 17, 18. Defant et al. We developed a model of opinion dynamics in which opinions are observed to converge to one average opinion and to clusters of opinions.19. Their model consists of the following populations: \(N\) Agent \(I\) with continued opinions \({x}_{i}\). At each timestep, two randomly chosen agents of hers “meet” and readjust their opinions if the magnitude of their disagreement is less than a threshold. \(\Valepsilon\). Suppose two agents have an opinion. \(X\) and \(x^{\prime}\) and that \(\left| {x – x^{\prime} } \right| < \varepsilon\)opinions are adjusted according to:
$$x = x + \mu \cdot \left( {x^{\prime} – x} \right)\;{\text{and}}\;x^{\prime} = x^{\prime} + \mu \cdot \left( {x^{\prime} – x} \right)$$
(1)
where \(\mu\) is the convergence parameter taken between 0 and 0.5 during the simulation. Defant et al.The value of \(\Valepsilon\) The main influence on the dynamics of the model, when it is high it causes convergence to one opinion and when it is low it causes polarization/fragmentation (convergence to multiple opinions).19. \(\mu\) and \(N\) It only affects the convergence time and the final opinion distribution. They applied their model to agents’ social networks. This allows an agent in the model to interact only with her four neighbors connected on the grid (so the random selection of agents to interact with is only from the connected neighbors). And got the same result.
Hegselman and Krauss developed a model with limited confidence to investigate opinion fragmentation, of which consensus and polarization are special cases.20. The Hegselmann-Krause (HK) model is defined as:
$$x_{i} \left( {t + 1} \right) = \left| { I\left( {i,x\left( t \right)} \right)} \right|^{ – 1} \mathop \sum \limits_{{j \in I\left( {i,x\left ( t \right)} \right)}} x_{j} \left( t \right)\;{\text{for }}\; t \in T$$
(2)
where \(I\left( {i,x} \right) = \{ 1 \le j \le n| \left|x_{i} – x_{j} \right| \le \varepsilon_{i} \}\ ) and \(\varepsilon_{i} \ge 0\) Agent trust level \(I\).Agent \(I\) only accept those agents \(j\) take into account that their opinion differs only slightly from your own \({\Valepsilon }_{i}\). The base case assumes a uniform confidence level. In other words, \({\Valepsilon }_{i}=\Valepsilon\) for all agents \(I\). The authors found that the higher the confidence threshold value, the more \(\Valepsilon\) Consensus is achieved, but low values ​​create polarization and fragmentation. In every run, the range of opinions decreases as the simulation is run, regardless of whether consensus or polarization occurs.
Fu et al.We modified the Hong Kong model by dividing the population into open-minded, moderate, and closed-minded agents.twenty one. They found that the number of final opinion clusters was dominated by closed-minded agents. Open-minded agents cannot contribute to consensus building, and the presence of open-minded agents may lead to diversification of final opinions.
Cheng and Yu suggested that individuals’ opinion formation and expression may differ in many social situations. This is because individuals feel pressured to express opinions similar to the public opinion of their in-group.twenty two. They propose a bounded trust plus group pressure model in which each individual forms an internal opinion about the limits of trust and expresses his or her opinion taking into account group pressure. In groups where all individuals face group pressure, consensus is always reached. In mixed groups with both pressured and non-pressured people, the consensus threshold ε is significantly lowered, and group pressure does not necessarily help promote consensus. Although similar to other models, polarization does not occur in their work.
More recently, Condie and Condie divided social influence into assimilating and differentiating types.2. Assimilative influence occurs when opinions converge toward a consensus or fragment into two or more converging clusters, all within the initial opinion. Differentiation effects, which are the focus of our study, occur when individuals with very different opinions can influence each other, leading to extreme disagreements (see Figure 1) . Condie and Condie proposed the Social Influence and Event Model (SIEM).2 It is built on the HK bounded trust model, \(\Valepsilon\) used as a confidence threshold, the main differences are: (1) agents form a social network; (2) The individual \(I\) Their opinion will only change if they are convinced. \({C}_{j,t}\in [\mathrm{0,1}]\)lower than the average certainty of other individuals interacting at once \(t\), (3) Most importantly, events can affect many individuals simultaneously within a limited period of time. Events can have a significant impact on the distribution of opinions. This is because an individual can only interact with a small number of other individuals at a given time, whereas its influence acts simultaneously on a large part of the population. The simulation results showed that the event-free SEIM exhibited a set of behaviors produced by other influence models under different levels of confidence thresholds. \(\Valepsilon\) leading to consensus (or assimilative influence in its definition).In the presence of strong events, the confidence threshold is \(\Valepsilon\) is high (low isomorphism), opinions oscillate between extremes, and confidence thresholds are \(\Valepsilon\) is low (homosexuality is high), and opinions are divided into two extremes.condi and condi2 measures were also introduced. conflict, \(\Delta {O}_{t}=SD({O}_{i,t})\)in a population, which they defined as the standard deviation of individual opinions. \({O}_{i,t}\) Entire population at timestep \(t\).
Developing these ideas further, Macy et al. Use a general model of public opinion dynamics to reverse the self-reinforcing dynamics of partisan polarization even under external threats such as a global pandemic or economic collapse. demonstrated the existence of a tipping point that may not be sufficient totwenty three. Agents in the model initially have random positions in a multidimensional issue space consisting of membership in one of two equal-sized parties and positions on 10 issues. Agents can then move closer to their immediate neighbors and away from those with whom they disagree, depending on the agent’s tolerance for disagreement and the strength of their party identity relative to their ideological commitment to the issue. Update your position on the issue. They manipulated the agent’s tolerance for disagreement and the strength of the parties’ identification, and introduced an exogenous shock in response to the event (following Condie and Condie)2) create common benefits against common threats (such as a global pandemic).
All of these studies demonstrate the value of this form of modeling for investigating opinion dynamics while assuming that expressed opinions are the same as actual opinions.