Abstract
Objective:
This article provides an overview of the characteristics of misinformation and information attack and their effects on the perceptions of the public, with the objective of outlining potential solutions and needed research for countering this growing problem.
Background:
Society is facing a significant challenge from the spread of misinformation through websites and social media that has driven a divergence in people’s perceptions and understanding of basic facts associated with many issues relevant to public policy decisions, including the economy, taxation, and the deficit; climate change and the environment; and vaccinations and public health and safety. A number of factors are driving this fracture, including information presentation challenges that lead to poor information understanding, deliberate information attacks, social network propagation, poor assessments of information reliability, and cognitive biases that lead to a rejection of information that conflicts with preexisting beliefs.
Results:
A framework for understanding information attack is provided, including common sources, features, avenues, cognitive mechanisms, and major challenges in overcoming information attacks.
Conclusion and Application:
Potential solutions and research needs are presented for improving people’s understanding of online information associated with a wide range of issues affecting public policy.
Introduction
Today’s information age is characterized by widespread Internet connectivity and global access to computers and mobile information devices. The digital revolution has markedly transformed society by allowing individuals easy and almost immediate access to information from a wide variety of sources on almost any topic imaginable.
Despite this prevalent and easy access to information, however, people face many challenges in developing an accurate understanding of the facts underlying public policy, which is critical for well-informed decision making. This problem is nonpartisan and includes an understanding of everything from the economy, government spending, jobs and taxes, to climate change, health care, vaccinations, genetically modified foods, and crime. Research shows that 81% of registered voters report disagreeing on the basic facts associated with policy decisions (Pew Research Center, 2016). There are many reasons for people to make different choices (including different priorities, worldviews or group identities), but it is more troubling that they are basing their decisions on very different perceptions of the facts. Without forming a common understanding of the ground truth underlying a problem, it is quite difficult, or often impossible, for even well meaning people to arrive at good solutions.
A key problem in the public policy space is that a significant amount of information may be deliberately incomplete, inaccurate, or misleading (Del Vicario et al., 2016; Lewandowsky, Ecker, Seifert, Schwarz, & Cook, 2012). In addition to the benign misinformation that litters the Internet, there has been a growth of deliberate information attacks—defined here as the intentional spread of false information to support the goals of the attacker. False information includes articles, reports, websites, or statements that are intentionally and verifiably false and likely to mislead recipients (Allcott & Gentzkow, 2017). It does not include small reporting errors (often quickly corrected), or information that is unflattering or in conflict with one’s worldview.
The central premise of this article is that information attacks are a significant threat, that they capitalize on certain human cognitive characteristics and weaknesses, and that the human factors profession can contribute to developing methods to overcome this new challenge. Although many information attacks may be directed against government or industry targets via cyber attacks, this article is specifically focused on understanding and combating information attacks in the public sphere.
Historically, human factors research and practice has been focused on improving people’s ability to perceive and process data in operational and organizational contexts. In addition, this knowledge can and should be applied to the public policy information space for several reasons. First, human factors solutions have often been focused on the display of data to support accurate perception and understanding (Endsley & Jones, 2012; Schneiderman & Plaisant, 2009; Tullis, 1983; Wickens, Hollands, Banbury, & Parasuraman, 2016). With the increased reliance on online information by the public, this research base can be easily applied to this domain. Secondly, while the human factors field has most often been directed at human interaction with technology in work settings, this is not universally the case. Human factors research is also applied in nonwork settings such as forensics (Laughery & Wogalter, 2008), computer games (McLaughlin, Gandy, Allaire, & Whitlock, 2012), children’s issues (Mathias & Brickman, 2015), and the design of voter ballots (Jastrzembski, 2004), for example. Lastly, to paraphrase Thomas Jefferson, a properly functioning society depends on an informed electorate; thus it seems prudent that the field direct its energies toward this important problem.
To promote an accurate understanding of online information, the human factors field must grapple with several challenges that are outlined in the information attack framework presented in Table 1. Aligned with this framework, I will first discuss the sources, features, and avenues of information attacks, followed by the basic human cognitive mechanisms that are at work in processing misinformation and that may be targeted by information attacks, as well as the significant human cognitive challenges that exist in overcoming misinformation. Based on this overview, I will then propose several areas where the human factors profession may offer potential solutions and where significantly more research is needed to combat information attacks.
Information Attack Characteristics
Characteristics of Information Attack
Sources of Information Attack
Political, internal
Misinformation can occur often for political reasons (e.g., a tactic by a politician or politically motivated group or individual), with the false message then repeated by other individuals, either knowingly or unknowingly (Lewandowsky et al., 2012). For example, politicians were a major source of misinformation during the 2009 healthcare debate, including claims that the proposed law included “death panels” (Lewandowsky et al., 2012). Inaccurate public policy information is widely distributed by a variety of individuals and political groups via websites, Twitter, and Facebook (Del Vicario et al., 2016; Ratkiewicz et al., 2011; Yin, Han, & Yu, 2008).
Political, external
In addition, the deliberate spread of misinformation or propaganda may be the strategy of a foreign government. Social media has been used to distribute misinformation in order to affect elections in multiple countries (Schåfer, Evert, & Heinrich, 2017). In the 2016 U.S. presidential elections, considerable evidence shows that the United States was subject to an extensive cyber information attack by Russia in a significant escalation of their efforts to attack U.S. international leadership (Office of the Director of National Intelligence, 2017).
Misinformation was a favored strategy of the former Soviet Union and remains an important part of Russian military doctrine today. “The Russian leadership invests significant resources in both foreign and domestic propaganda and places a premium on transmitting what it views as consistent, self-reinforcing narratives regarding its desires and redlines, whether on Ukraine, Syria, or relations with the United States” (Office of the Director of National Intelligence, 2017, p. i). Although the prevention of cyber attacks on government and military systems have received serious attention (Lynn, 2010), these events show that information attacks in the public opinion arena are a growing threat that must be addressed.
Commercial
Other misinformation is motivated primarily by monetary gain, including the promotion of corporate or other vested interests (Allcott & Gentzkow, 2017; Lewandowsky et al., 2012). This includes a misinformation campaign by the tobacco industry on the dangers of cigarette smoking (Smith et al., 2011), a campaign by the petroleum industry to cast doubt on the science of climate change (Hoggan, 2009; van den Hove, Le Menestrel, & de Bettignies, 2002), and an effort by Greenpeace to discredit genetically modified foods (Parrott, 2010). There has also been a significant increase in websites featuring completely false stories that are developed with the goal of increasing advertising sales via a high volume of web traffic (Allcott & Gentzkow, 2017; Chen, Conroy, & Rubin, 2015).
In some cases there may be an overlap between these sources of information attacks (i.e., coordination between commercial and political sources, or between internal and external political sources).
Features of Information Attacks
Information attacks can take multiple approaches. Of note, the information attacks by Russia on the 2016 U.S. elections involved multipronged efforts, including coordinated actions by Russian government agencies, state-funded media, third-party intermediaries, and paid social media users or trolls, along with cyber attacks (Office of the Director of National Intelligence, 2017). This approach has been characterized as involving high-volume, rapid, continuous and repetitive attacks that put forth early (although erroneous) information that favors their agenda, which is then repeated through multiple channels, creating familiarity and anchoring effects that color receipt of later information (Paul & Matthews, 2016). It may include completely false information as well as information that combines factual and manufactured data, such as doctored photographs and actors hired to portray victims in fake reporting (Paul & Matthews, 2016).
Other information attacks have spread numerous false stories in an effort to confuse people or to find a story that sounds convincing, as was the case in the downing of Malaysia Airlines Flight 17 (Paul & Matthews, 2016). Many propaganda efforts include both false information and the provision of selective information (i.e., cherry picking) designed to create a false understanding of events (O’Connor & Weatherall, 2018).
Avenues of Information Attack
Media and advertising
Information attack can be directed through both legitimate and phony news sites, including print, online, and broadcasting news sources. Online news websites have significantly expanded the number of potential news sources available, with varying levels of veracity which can fool many unwary readers (Pew Research Center, 2018). In an echo effect, misinformation provided through less reliable sources may be picked up and repeated by legitimate news sources, adding to the dissemination of false news stories (Paul & Matthews, 2016). Information attacks may involve paid advertising or may be part of the actual media content, with lines between them blurred or often invisible.
Individuals
People also play a role in propagating false information. Some 23% of people report sharing false news, either knowingly or unknowingly (Barthel, Mitchell, & Holcomb, 2016). False news stories are 70% more likely to be retweeted by people on Twitter than regular news stories (Vosoughi, Roy, & Aral, 2018).
Social media
Social media often exacerbates misinformation campaigns, acting to rapidly propagate and magnify false stories placed in news articles or websites. False news stories are accessed through social media 42% of the time, whereas actual news sites are accessed only 10% of the time (Allcott & Gentzkow, 2017). False news (as determined by six fact-checking organizations) was also shown to spread more rapidly and much further on Twitter than true information, most likely due to its novelty (Vosoughi et al., 2018). Users have been found to select and share information online around specific topics within so-called communities of interest, which provides reinforcement and increases segregation and polarization (Del Vicario et al., 2016).
Automated network propagation
Although the use of propaganda in politics and warfare has a long history, its more recent incarnation as computational propaganda, which uses automated bots and social media, has provided a significant boost to its spread (Bolsover & Howard, 2017). It is estimated that between 9% and 15% of Twitter accounts are bots and 60 million bots are active on Facebook; these accounts were highly active in creating political content during the 2016 U.S. presidential election and 2017 French presidential election (Lazer et al., 2018). Bot traffic appears to be particularly adept at propagating and multiplying specific posts through bandwagoning on existing hashtags and themes (Schåfer et al., 2017).
Cognitive Processes and Mechanisms Targeted by Information Attack
Projections of the effect of different policy decisions are highly dependent on the individual’s situation representation (i.e., their comprehension of the state of affairs regarding a particular topic). Figure 1 shows the Endsley (1995) model of situation awareness (SA) annotated to show several key cognitive features that can be the targets of information attack negtively impacting these situation representations.
Cognitive mechanisms affected by information attack.
Attention
In a world with many competing pieces of information, that with the most salient characteristics often grabs an individual’s attention. Online information attacks often utilize salient cues, such as outrageous or incendiary headlines, as “click bait.” Information that is easy to access (such as that appearing in e-mail or ads on frequented websites) is more likely to be attended to, as well as that from known associates.
Endsley and Jones (2001) identify several key methods for degrading SA through information attacks targeted at the attention process, including too much information, too fast a flow of information, disorganized information content, and dissonant information, which all work to disrupt information processing efficiency. This is a perfect description of the state of online information. A majority (58%) of Americans say it is harder to be informed today due to the plethora of information and news sources available (Gallup-Knight Foundation, 2018). Endsley and Jones (2001) also describe attacks on information prioritization. For example, information attacks may be used to redirect people’s attention toward certain stories and away from others (i.e., provide a distraction in line with the attacker’s goals).
Anchoring
Early information acts to establish or direct the selection of a mental model for understanding and gathering additional data. This creates an “anchor” or primacy effect in which early information is often given more weight and can have a significant effect on what information is later attended to or believed (Kahneman, Slovic, & Tversky, 1982). If early information is erroneous, this can have a significant negative effect on later information processing.
Confirmation bias
Although individuals may make decisions based on only one story or piece of information, often they will continue to acquire additional information. Goals and goal-directed processing can drive the search for relevant news stories, which often acts to confirm preexisting expectations or beliefs held in the mental model (Kahneman et al., 1982; Nickerson, 1998). When those beliefs or expectations are incorrect, confirmation bias influences the individual’s search for only confirmatory information and the neglect or discounting of conflicting evidence (Nickerson, 1998).
Perception and confidence level
Key information about a story will be perceived, along with a level of confidence (trust) in the information. How much confidence to place in information is a significant component of SA (Endsley, 1995). Confidence in information received is typically determined based on the reliability of its source, the presence of incongruent or conflicting data, missing information, the timeliness or latency associated with the data, and the presence of noisy or ambiguous signals (Endsley & Jones, 2012). When different data sources are in agreement, this can act to increase confidence levels (Gilson, Mouloua, Graft, & McDonald, 2001). Conversely, conflicting data can lead to lower levels of confidence and slow down decision making (Bolstad & Endsley, 1999).
Endsley and Jones (2001) point to information attacks directed at information confidence, which significantly slows information processing and delays decision making, even when accurate information is present. If decision makers are uncertain or misled about the veracity of online information, it significantly disrupts their ability to process, integrate, and act on that information. Kahneman (2013) describes how uncertainty can shift a person from rapid “system 1” type thinking toward slower, more analytical “system 2” type processes.
Comprehension and mental models
Ultimately, individuals must interpret new information together with other preexisting knowledge to form an internally coherent understanding of an issue. When new information is in agreement with already existing mental models, this is fairly effortless. When it is in conflict, people are more likely to discount the new information or explain it away in order to retain the existing (although possibly incorrect) mental model (Nickerson, 1998).
Attacks on information interpretation affect which information is attended to and how it is integrated with existing information (Endsley & Jones, 2001). By creating preexisting, but incorrect, mental models of a subject, it is possible to induce people to ignore conflicting information and seek out confirmatory information based on anchoring and confirmation biases (Kahneman et al., 1982). In one study, we found that 66% of people failed to notice or explained away information that did not fit with a mental model established by early information, even when the new information was in significant disagreement (Jones & Endsley, 2000). People find ways to explain away conflicting data that disagrees with their existing mental model, either consciously or unconsciously. This cognitive dissonance can be quite strong and difficult to overcome (Festinger, 1962).
Projection, goals, and motivation
Information attacks can also be directed at creating inaccurate projections of the likely future (e.g., whether global warming will continue), and the projected effects of policy decisions (e.g., whether tax cuts will result in job growth or increased federal debt).
In most operational systems (e.g., aircraft piloting, air traffic control, military command and control), the cost of inaccurate situation awareness is high, and it has generally been assumed that people strive to maintain as accurate a picture of the situation as possible. In the public policy arena, however, this may not be an accurate assumption. People may instead be more motivated to prove their own worldview correct, leading to the neglect of conflicting information, or the interpretation of news stories to fit with an existing mental model. Flynn, Nyhan, and Reifler (2017) provide considerable evidence that directionally motivated reasoning (that which seeks out information in agreement with one’s political views and counterargues that which does not) is pervasive and due to a strong affective response or identity threat. Subtle information attacks may manipulate people’s motivational frameworks, pushing them toward defensive mentalities, for example, that exacerbate directionally motivated reasoning.
Challenges for Combating Information Attacks
Poor understanding of information reliability
People can have trouble in determining what online information to believe. Allcott and Gentzkow (2017) found 8% of people reported seeing and believing false stories. One poll found that only 27% of Americans are very confident that they can tell when a news source is reporting factual news versus commentary or opinion (Gallup-Knight Foundation, 2018). People also have been found to be fairly poor at determining whether photographs have been altered, creating additional problems with judging the veracity of online imagery (Nightingale, Wade, & Watson, 2017).
Conversely, people may reject accurate stories as false. The Gallup-Knight foundation (2018) found that 42% of Republicans and 17 % of Democrats consider accurate news stories that portray a politician or political group in a negative light to always be false. Without the ability to accurately assess which online information to believe, many people will incorrectly dismiss information that does not agree with their preexisting beliefs (Cook, Ecker, & Lewandowsky, 2015).
The challenge of accurately knowing how much confidence to place in online information sources has been made more complicated by attacks on the media and fact-checking sites (Shin & Thorson, 2017). By attempting to negate the legitimacy of fact-checking sites and accurate news sources, a circular chain is created in which party loyalists do not consider negative information credible, thus those within the party can conduct actions with impunity. Allcott and Gentzkow (2017) found that older people, those with more education, and those who spend more time consuming news are the most accurate at assessing which news is correct or false.
Social reinforcement
Shared realities or ideologies often occur within interpersonal relationships and social groups, which forms the basic mental model for interpreting information (Jost, Ledgerwood, & Hardin, 2007). Evidence suggests that people’s political judgments are largely influenced by cultural group identity, independent of their attitudes about particular topics (Wlezien & Miller, 1997). Multiple studies demonstrate that attitudes toward public policies are highly driven by the position of one’s political party, much more so than the policy content or the individual’s ideology, primarily due to shifts in confidence about the factuality of a policy and its perceived moral implications (Cohen, 2003). This often occurs subconsciously, and people deny they have been so influenced (Cohen, 2003). Additionally people tend to overweight the reliability of information they receive from people they know (Seifert, 2002), a significant problem when social media are used to spread misinformation.
Resistance to conflicting information
Significant numbers of people believe many false news stories and will often cling to false beliefs once established, even in the face of considerable evidence to the contrary (Allcott & Gentzkow, 2017). These persistent false beliefs occur on both ends of the political spectrum and include topics such as climate change, the effects of vaccinations, election tampering, the presence of weapons of mass destruction in Iraq, and complicity in the 9/11 attacks.
Considerable research has examined the “continued influence effect,” in which early misinformation continues to be believed even after retractions and corrections of that misinformation (see Lewandowsky et al., 2012, for a review). Initial acceptance of new information is often automatic (unless people are motivated to more thoughtfully evaluate its veracity through a consideration of its compatibility with other known information, internal consistency and plausibility, source credibility, and degree of belief by other people). If original information is later retracted or corrected (e.g., new information contradicts the early information), a residual effect occurs in which many people (generally at least half) will continue to remember and believe the original misinformation.
The continued influence effect occurs due to the establishment of a mental model based on initial information, which is resistant to change because it is easier to retain an incorrect mental model than an incomplete or conflicting one (Johnson & Seifert, 1994; Wilkes & Leatherbarrow, 1988). An alternate explanation is that there is a failure of controlled memory processes, wherein people confuse which information was correct with which information was incorrect, or in which the incorrect information competes with the corrected information during retrieval (Ayers & Reder, 1998; Moscovitch & Melo, 1997). The third common reason provided for the continued influence effect is fluency or familiarity (Jacoby, 1999). Early misinformation continues to exert influence over time due to its increased perceived familiarity (which tends to get inadvertently reinforced during correction efforts) and coherence with related material. Myth-versus-fact presentations can accidentally act to increase beliefs in the myths being debunked by increasing perceived familiarity due to reinforcement, for example (Jacoby, 1999; Skurnik, Yoon, Park, & Schwarz, 2005).
The continued influence effect is strongest for information that is consistent with preexisting worldviews and for older adults (Lewandowsky et al., 2012). Allcott and Gentzkow (2017) found that people are 15% more likely to believe ideologically consistent false headlines, and this effect is strongest for those with ideologically segregated social media networks. Other research argues for a tipping point at which extensive disconfirming information will finally break through such ideological resistance, as needed to resolve built up anxiety over conflicting information (Redlawsk, Civettini, & Emmerson, 2010).
Backfire effect
Correct information stating there were no weapons of mass destruction (WMDs) in Iraq was only effective for Democrats in dispelling incorrect beliefs, but created an actual increase in the number of Republicans who believed that WMDs were present (Nyhan & Reifler, 2010). Similar effects have been found with efforts seeking to discount the effect of vaccines on autism, where parents with the least favorable views toward vaccines become even less inclined to vaccinate after viewing debunking messages (Nyhan, Reifler, Richey, & Freed, 2014).
These examples demonstrate that efforts to correct false information can accidentally act to reinforce it instead in what has been labeled a boomerang or backfire effect (Byrne & Hart, 2009; Skurnik et al., 2005). People can end up even more likely to believe older, incorrect information following correction efforts. This backfire effect occurs due to the continued influence of older information and because people may actually reinforce their prior beliefs as they act to argue against conflicting information.
Poor feedback
Feedback loops in the public policy arena are often slow and noisy. It may take years to determine the effect that a particular policy has on outcomes of interest (e.g., did a tax cut lead to significant increases in employment paying for itself, or to increases in the deficit?), and even then information on those effects is often buried in confusing statistics, clouded by other events and actions, and possibly distorted by partisan reporting. Thus, it may be very difficult or slow for incorrect mental models or beliefs to get corrected by decision outcomes. The challenges for creating accurate SA in the public arena are significant in ways that may not be typical in other domains.
Cognitive Engineering Solutions and Research Needs
Although these challenges may seem quite difficult to overcome, there are several potential avenues in which human factors research can contribute to addressing information attack, including improving information presentation, confidence, and integration; addressing goals, motivations, and mental models used in processing information; overcoming confirmation bias and social forces; and managing computational propaganda. In some cases, relevant research exists; however, overall far more research is needed to address this growing concern.
Information Presentation
There are a number of factors associated with information presentation that can improve the accuracy of people’s understanding of world events.
Immediate and repeated communications of facts
To avoid the familiarity backfire effect and avoid reinforcing false information, messages should focus on the facts to be communicated rather than the misinformation to be corrected (Lewandowsky et al., 2012). Since early information is given more weight by creating an anchoring effect, it is important that misinformation be corrected swiftly and repeatedly to help overcome this challenge, and that misinformation not be reinforced by repeating it in the correction.
Simplicity
Information that is easier to process is more likely to be accepted (Schwarz, Sanna, Skurnik, & Yoon, 2007). A few clear simple arguments are more likely to be accepted than a long, complex one, and clear, simple information presentation is important. Improved communications on the scientific evidence behind climate change has been advocated, for example (Somerville & Hassol, 2011).
Framing
Information framing has also been shown to be effective at avoiding backfire effects. For instance, Republicans are more likely to accept a “carbon offset” than a “carbon tax” because the word tax conflicts with their established worldview (Hardisty, Johnson, & Weber, 2010). Focusing on the dangers of communicable diseases has been found to increase positive attitudes toward vaccinations (Horne, Powell, Hummel, & Holyoak, 2015), whereas efforts to dispel myths surrounding vaccines have been largely unsuccessful (Nyhan et al., 2014).
Warning flags
Advanced warnings that information may not be true can minimize the effect of false information by reducing the level of encoding of the information (Lewandowsky et al., 2012). This can include explicit warnings in the media when reporting on known false statements by others, or warnings on news distribution sites regarding suspect articles. For example, when the warning “some politically motivated groups use misleading tactics to try to convince the public that there is a lot of disagreement among scientists” was provided along with correct information, the effect of misinformation about climate change was substantially reduced (van der Linden, Leiserowitz, Rosenthal, & Maibach, 2017).
Information Confidence
More support is needed for assessments of the accuracy of information. Encouraging skepticism of the source of misinformation has been found to reduce its influence, for example (Lewandowsky et al., 2012). Refutations of false information from unexpected sources, particularly when it comes from inside one’s own group or when it runs counter to partisan predilections, are most effective (Lewandowsky et al., 2012). For example, Berinsky (2017) found that efforts to debunk the “death panels” claim associated with the Affordable Care Act were most effective when they came from a Republican.
Far more research is needed in this area. New methods are needed to help people assess the reliability of online sources, including websites, e-mails, and information spread through social media. Direct input of useful data from relevant fact-checking sites and data on the reliability of different media or websites may be useful, for example. Given that there can be many different news stories or websites on topics, tools that help users to manage information from sources with different levels of reliability may be useful. There also may be a benefit to helping people understand whether additional information is new (adding to credibility through confirmation), or whether it stems from the same initial source as previously received information (i.e., repetition that does not add to reliability).
Information Integration
The basic challenge of the ongoing news cycle is that information comes fast, is disorganized (coming from different sources), and may often contain dissonant information that is hard to process. Although some news stories include graphics, most of it is highly verbal in nature, and people are left to mentally integrate and interpret all of it, often while they are busy with tasks of daily life. Websites featuring integrated information in graphical formats can go a long way toward helping people understand complex subjects. Effective formats for helping people to better understand the interacting factors in these complex issues and how various policy approaches impact them would be highly beneficial.
Graphics have been shown to be effective in reducing misconceptions (Pandey, Manivannan, Nov, Satterthwaite, & Bertini, 2014). They provide more clarity and promote the development of a more nuanced mental model. For example, when different sources of information both promote and deny climate change, individuals are left with uncertainty or simply choosing one set of sources over another, most likely those they already identify with and trust. But if relevant and accurate data are shown graphically in a clear manner—with relevant contextual information—it allows the viewer to draw his or her own conclusions with a higher level of confidence.
Significant research has been conducted in the human factors field on how to present graphical information (e.g., Gillan, Wickens, Hollands, & Carswell, 1998), and graphically presented data has been shown to decrease confirmation bias in analyzing new information (Cook & Smallman, 2008). New research is needed on how to best utilize graphics to inform an accurate understanding of public policy issues.
While graphs on a variety of topics are available online, many are ineffective because they lack the relevant context for helping people to understand them. Effective graphics need to include relevant events (e.g., the passage of relevant laws or policy changes, key events, responsible administrations or office holders) so that data can be properly understood in context.
For example, in addition to changes in unemployment rates over time, Figure 2 shows how it has changed across Republican and Democratic administrations and is annotated with key information for interpretation. The source of the data is included to support confidence assessment (Bureau of Labor Statistics, 2018). Following the recommendations of Cook and Lewandowsky (2011), a summary of the findings and narrative is provided to explain why the data go against common political assumptions. The effectiveness of such a presentation needs to be assessed and compared with alternate presentation approaches to determine the most important features to include.
Annual unemployment rate since 1969.
It should be recognized that many people have trouble understanding graphs (Galesic & Garcia-Retamaro, 2011). More research is needed on effective methods for overcoming these challenges via new graphical presentation techniques, multimedia tools, or combinations of graphs and text. In addition, new forms of media presentation are needed to counter cherry picking of information by showing the broader range of relevant information, placing information into the appropriate context for accurate understanding. These approaches should take advantage of the information persistence that is possible with websites that is not present with more transient new stories.
Goals and Motivation
It may be very difficult to address subconscious motivations that people bring with them. Lewandowsky et al, (2012) advocate that because people are more likely to reject information that conflicts with their worldview, particularly those who are most fixed in those views, it is best to simply direct correction efforts at people who are less entrenched or undecided. Other research points to the effects of emotional states on information acceptance. For example, messages that promote self-affirmation have been found to increase the likelihood that conflicting information will be accepted (Cohen, Sherman, Bastardi, Hsu, & McGoey, 2007). Research is needed to find effective ways to reduce defensive mindsets and to increase objective information processing over directed reasoning. For example, messaging that acts to counter the “us versus them” and group identity alienation of certain political communications should be explored.
Mental Models
Because people have difficulty in fitting conflicting information into their existing mental models, leading to rejection of new data, they need a new narrative that allows for a coherent internal model (Seifert, 2002). Cook, Lewandowsky, and Ecker (2017) showed that exposing the logical flaws in anti-climate-change messages was effective, as was emphasizing the scientific consensus on climate change, for example. Alternate narratives create a foundation for new information and may help explain old information in a new light, or may provide reasons why a misinformer promoted false information.
The use of integrated information, presented graphically, can be highly useful in building more accurate mental models of relevant topics. Interactive simulation tools that allow people to better understand the cause and effect of interacting variables have also been found beneficial (Berger, 2001; Sterman, 2006). More research is needed on the development of effective simulations that can create more accurate mental models and overcome preexisting biases. These tools should be directed at helping a lay audience rapidly develop a better understanding of topics and a better ability to project the effects of different policy decisions.
Overcoming Confirmation Bias
Some research has focused on how to reduce confirmation bias and increase the likelihood that people will process information openly and objectively. Studies show that encouragement to form accurate opinions (valuing accuracy over directional reasoning), for example, can reduce or eliminate preexisting biases toward directional reasoning (Flynn et al., 2017; Taber & Lodge, 2006). Information that comes from party elite or in-group members tends to be more accepted and can reduce confirmation bias (Flynn et al., 2017). Lehner, Adelman, Cheikes, and Brown (2008) found that confirmation bias affected the weighting placed on confirming over disconfirming information, but that training people to analyze competing hypotheses was successful for only some people.
Other research has found an open-minded thinking characteristic that has a significant effect on promoting analytic thinking as opposed to directional reasoning (Haran, Ritov, & Mellers, 2013; Stanovich & West, 1997). Research is needed to determine whether this characteristic can be trained or fostered by other interventions. Far more research is needed on effective methods for reducing confirmation bias and cognitive dissonance and encouraging open-minded, analytic processing of online information.
Overcoming Social Forces
Given the strong role of one’s social group in opinion formation, research on reducing group think and changing opinions in group settings needs more emphasis. Some research has shown that political messaging can be framed in ways that are consistent with group values, for instance, increasing their acceptance and effectiveness on even controversial issues (Westen, 2008).
Managing Computational Propaganda
Social media tools as well as news aggregation sites have become important daily information sources for many people, but, as such, face a significant challenge in addressing the many misinformation attacks that are directed through them. Tools and approaches are needed to directly address automated bots, false news, and other misinformation that is being routed through these platforms. This has been recognized as a sociotechnical problem (Bolsover & Howard, 2017).
Many social media companies are trying to address this challenge. YouTube has announced it will be flagging content funded by state-backed media (YouTube, 2018). Facebook has acted to reduce the distribution of disputed stories by providing links to fact-checking sites. (Lyons, 2017). Some research found showing related articles was successful at reducing misperceptions by providing better context for understanding information and avoiding backfire effects (Bode & Vraga, 2015).
Ozturk, Li, and Sakamoto (2015) tested methods for countering false information on Twitter. They found that providing counterinformation to the false message was somewhat effective at reducing retweets, as well as noting that the false information had been flagged on fact-checking websites. Kumar and Geethakumari (2014) demonstrated the ability to automatically determine misinformation on Twitter using a collaborative filtering approach with a 90% accuracy, which may lead to the ability to conduct real-time correction flagging. Amazeen Thorson, Muddiman, and Graves (2018) investigated the utility of graphical fact meters on fact-checking sites and found them effective for evaluating nonpolitical information, but were not more successful than content only for political information.
More research is needed to determine the effectiveness of these strategies and to determine the best methods for overcoming misinformation in social media to allow users to better calibrate the reliability of information provided. Methods that introduce screening tools or negative weighting for objectively false information are needed, as well as those that reduce the effort required by users to assess information accuracy.
Conclusions
The spread of information attacks in the public policy domain has become a significant problem challenging society. Combating this challenge presents a unique opportunity for the human factors profession to investigate and validate innovative solutions. Just as accurate situation awareness is important for the operation of aircraft and power plants, it is equally important for good decision making in the electorate.
The information attack framework presented here outlines many of the key features of this problem and points the way to some potential solutions. However, far more research is needed to overcome the many cognitive difficulties associated with information attacks. Will our research community rise to this challenge?
Key Points
Misinformation, propaganda, and information attack are acting to undermine the democratic process by reducing the public’s understanding of the facts associated with policy decisions.
A framework is presented for understanding information attack including sources, features, and avenues of information attacks, human cognitive mechanisms affecting processing of misinformation, and cognitive challenges in overcoming information attack.
Human factors research can be applied to help address this problem by improving information presentation, confidence, and integration; addressing goals, motivations, and mental models used in processing information; overcoming confirmation bias and social forces; and managing computational propaganda.
Footnotes
Mica R. Endsley is President of SA Technologies and is the former Chief Scientist of the U.S. Air Force. She received a PhD in industrial and systems engineering from the University of Southern California in 1990. She has published extensively on situation awareness, automation, and system design.