Analyzing Misinformation Through The Lens of Systems Thinking Umme Ammara Information Technology University / Lahore, Pakistan [email protected] Hassan Bukhari Information Technology University / Lahore, Pakistan [email protected] Junaid Qadir Information Technology University / Lahore, Pakistan [email protected] Abstract than the traditional linear thinking of cause and ef- Recently, there has been an alarming rise fect. It posits that events and trends observed in a in the presence of false information online, system is inherent to its internal structure, which is which may be spread knowingly with mal- defined by the relationships between its elements intent (“disinformation”) or naively without and which gives rise to its purpose (John, 2000). knowing (“misinformation”). The spread of Applying the lens of systems thinking will con- both these types of false information disin- tribute in improving truth and trust issues; by en- tegrates public trust on the information ecol- abling us to focus on the interconnections of this ogy and can have deleterious effect on soci- complex problem. It will allow us identify root ety (e.g., when wrong health information is adopted or when polarization and hostility in- causes and anticipate long term consequences and creases due to incorrect news about others). unanticipated side effects of solutions under con- Various interventions have been tried (e.g., en- sideration. hancing digital literacy) with mixed results. There is a plethora of information on online me- Current analytical methods are not well suited dia in this era. Recently, there has been an alarm- for studying such complex adaptive systems as ing rise in the presence of false information on- they fail to capture the interactions and inter- line, which may be spread knowingly with malin- dependencies between the various parts of the system. In this paper, we propose the use of tent (“disinformation”) or naively without know- system thinking for modeling misinformation ing (“misinformation”). In this paper, we refer (for convenience and simplicity of exposition, to all kinds of false information online as misin- we refer to all false information as misinfor- formation for convenience and ease of exposition mation regardless of the intent). We demon- even though technically misinformation is differ- strate the use of system thinking tools such as ent from disinformation in that there is a definite causal loop diagrams and stock-and-flow mod- intention in the latter to harm and deceive unlike els for modeling misinformation. To the best of our limited knowledge, this is the first of its the former case. However, for the purpose of the kind of research regarding viewing misinfor- paper, we will misinformation to be synonymous mation from the perspective of system think- and inclusive of disinformation. The spread of ing. both these types of false information disintegrates public trust on the information ecology and en- 1 Introduction ables an infodemic that strains the ability of peo- How can we solve the problem of misinformation ple to find relevant authentic information from the and disinformation in an effective manner? Why Internet. do our problems persist despite the best efforts, Unfortunately the problem of taming the info- and why do our solutions to the pressing 21st cen- demic and countering misinformation is far from tury challenges in turn often perpetuate these prob- easy. This is because the modern information lem even further. These questions motivate us to ecology—comprising billions of humans who use view the rising issue of misinformation from the information sources managed by public and pri- perspective of systems thinking. System thinking vate owners and influenced by various regulations can translate a problem, such as prevalent misin- and policies using new media such as the Internet, formation, into a set of feedback loops that rep- social media, and mobile phone apps—is a com- resent reinforcing and balancing processes, rather plex adaptive social system. Such systems are well 55 Proceedings of the 2020 Truth and Trust Online (TTO 2020), pages 55–63, Virtual, October 16-17, 2020. known to be counterintuitive due to the presence have the way they do” (Meadows, 1999). Unfor- of various intertwined feedback loops that interact tunately, frequently various solutions to the prob- nonlinearly. This results in an overall system in lem of misinformation exacerbate the problem— which causes and their effect are linked together the cure being worse than the disease. As said by indirectly through circuitous paths, distant in time Lewis Thomas, “If you want to fix something you and space, which makes the job of analyzing the are first obliged to understand... the whole sys- efficacy of an intervention difficult and error-prone tem.” He further said even when one is anxious to (Forrester, 1971). Despite efforts to regulate mis- solve any problem of significant complexity, it is information through legislation, legal warnings to inappropriate to intervene naively and have ‘much corporations, and innovations in Artificial Intelli- hope of helping’ (Stroh, 2015a). Motivated by this gence (AI) to track and label cases of misinforma- viewpoint, this paper builds on selected concepts tion, authorities have failed to overcome the chal- discussed in misinformation literature via the lens lenges posed by misinformation (M.West, 2017). of systems thinking so that we can understand mis- Systems thinking is ideally suited to such com- information fully and bring sustainable and lasting plex adaptive systems. Systems thinking has been improvements while avoiding common pitfalls. defined as “the ability to understand the systemic 2 Primer on System Thinking interconnections in such a way as to achieve a de- sired purpose” (Stroh, 2015b). Systems thinking is Systems thinking had its genesis during the devel- a highly-developed discipline with many schools opment of the intellectual discipline of system dy- of thought (including system dynamics, complex- namics, which emerged in 1950 at MIT, USA. It ity theory, general systems theory, human system describes the bigger picture, so sustainable solu- dynamics, etc. (Stroh, 2015b)(Arnold and Wade, tions can be implemented with fewer resources. In 2015)) and highly-developed qualitative and quan- terms of looking at interconnected causes of com- titative tools (Kim, 1995) (some of which we dis- plex problems, addressing the underlying prob- cuss in the next section) (Stroh, 2015b)). lems, not symptoms, controlling the solution of The following four key distinctive thinking pat- the dynamic problem by changing system’s be- terns are said to distinguish systems thinking from haviour, thinking about the endogenous variables conventional thinking (Richardson, 2011): firstly, operating in feedback loops, discouraging policies the ability to think dynamically (e.g., using graphs that intend to foresee long term success based on over time); secondly, to think causally through quick short-term fixes and focusing on improv- feedback loops; thirdly, to think of stocks and ing the relationship among parts by coordinated flows (i.e., accumulation and transfer); and finally, changes over a long period system thinking dif- to think more deeply about endogenous influences fers from the conventional linear thinking (Stroh, (where the system itself is the cause of the ob- 2015a). served problems). System thinking can also be 2.1 System Thinking Tools understood by contrasting it with open-loop based Since the genesis of system thinking at MIT [as conventional thinking (see Table 1), which fails systems dynamics] in the 1950s, the system think- to take into account that social systems are more ing community has developed versatile set of properly modeled as multi-loop nonlinear feed- qualitative and quantitative tools such as causal back systems and in such systems hardly anything loop diagrams, graphical functions diagrams us- is ever influenced linearly in just one direction ing stock and flow models. These system think- (Forrester, 1971) and in nonlinear systems, “the act of playing the game has a way of changing the ing tools have been widely applied in various sec- tors such as healthcare, education, and manage- rules”. ment to gain insights into effective policy-making Systems thinking can allow us to rigorously an- and to study the behavior of complex systems. We alyze the anticipated outcomes of interventions discuss causal loop diagrams and stock and flow designed to mitigate misinformation. A pioneer models next. of the system thinking, Jay Forrester, describes the discipline as “take the knowledge we already 2.1.1 Causal Loop Diagrams have about details in the world around us and to Causal Loop Diagrams (CLD) present non-linear show why our social and physical systems be- cause and effect relationship in a simplified illus- 56 Table 1: Conventional vs. Systems Thinking [Credit: (Yaqoob et al., 2018); Details: (Senge, 2006)(Stroh, 2015b)] Conventional Thinking Systems Thinking Model of thinking Linear, causal, open-looped, immediate Nonlinear, multi-causal, close-looped with de- feedback layed feedback Determining a problem’s cause Obvious and easy to trace Indirect and non-obvious Cause of problems External to the system Internal (System-as-a-cause thinking) How to optimize? By optimizing the parts By optimizing relationships among the parts Where to intervene? Aggressive use of “obvious” solutions Careful change applied at the “leverage points” How to resolve problems? Cure the symptoms