Influencing the Others’ Minds: An Experimental Evaluation of the Use and Efficacy of Fallacious-Reducible Arguments in Web and Mobile Technologies

Influencing the Others’ Minds: An Experimental Evaluation of the Use and Efficacy of Fallacious-Reducible Arguments in Web and Mobile Technologies

PsychNology Journal, 2014 Volume 12, Number 3, 87 – 105

87

Influencing the Others’ Minds: An Experimental Evaluation of the Use and Efficacy of Fallacious-Reducible Arguments

in Web and Mobile Technologies

Antonio Lieto∗1 and Fabiana Vernero∗2

1Università degli Studi di Torino ICAR – CNR, Palermo

(Italy)

2 Università degli Studi di Torino

(Italy)

ABSTRACT

The research in Human Computer Interaction (HCI) has nowadays extended its attention to the study of persuasive technologies. Following this line of research, in this paper we focus on websites and mobile applications in the e-commerce domain. In particular, we take them as an evident example of persuasive technologies. Starting from the hypothesis that there is a strong connection between logical fallacies, i.e., forms of reasoning which are logically invalid but psychologically persuasive, and some common persuasion strategies adopted within these technological artifacts, we carried out a survey on a sample of 175 websites and 101 mobile applications. This survey was aimed at empirically evaluating the significance of this connection by detecting the use of persuasion techniques, based on logical fallacies, in existing websites and mobile apps. In addition, with the goal of assessing the effectiveness of different fallacy-based persuasion techniques, we performed an empirical evaluation where participants interacted with a persuasive (fallacy-based) and with a non-persuasive version of an e-commerce website. Our results show that fallacy-based persuasion strategies are extensively used in existing digital artifacts, and that they are actually effective in influencing users’ behavior, with strategies based on visual salience manipulation (accent fallacy) being both the most popular and the most effective ones.

Keywords: Mobile persuasion, web persuasion, logical fallacies, captology, e-commerce.

Paper Received 30/12/2014; accepted 31/12/2014.

1. Introduction

In the last decade several studies in the field of Human-Computer Interaction (HCI)

have started to focus attention on forms of persuasive interaction where the goal of

one of the two agents involved in the process, namely, the technological artifact, is that

of “orienting” the attitudes and/or behaviors of the other agent (the user) according to a *Corresponding Authors: Antonio Lieto Università degli Studi di Torino, c.so Svizzera 185, 10149, Torino E-mail: alieto@acm.org Cite as: Lieto, A., & Vernero, F. (year). Influencing the Others’ Minds: An Experimental Evaluation of the Use and Efficacy of Fallacious-Reducible Arguments in Web and Mobile Technologies. PsychNology Journal, 12(3), 87 – 105. Retrieved [month] [day], [year], from www.psychnology.org.

A. Lieto and F.Vernero

88

predefined direction. It has long been pointed out that digital artifacts which are

perceived as social actors, i.e., which elicit social responses and emotional

involvement on the part of their users, can apply similar persuasion strategies to those

used in human-human interaction (Fogg, 2003). Similarities in the arguments used by

persuasive technologies and human persuaders, however, are not limited to the cases

where computers are perceived as almost-living entities. In this paper, we start from

the hypothesis, first formulated in our previous work (Lieto and Vernero, 2013), that

there is a strong connection between logical fallacies (forms of reasoning which are

logically invalid but cognitively effective, studied since the antiquity in the fields of

logics and rethorics) and some of the most common persuasion strategies adopted

within digital technologies. We present the results of two studies: a survey carried out

in the e-commerce domain, aimed at ascertaining the use of persuasion techniques

based on logical fallacies in existing websites and mobile apps, and an empirical

evaluation where we compare user choices in a persuasive (fallacy-based) and in a

non-persuasive website, with the goal of assessing the effectiveness of different

fallacy-based persuasion techniques. The rest of the paper is organized as follows:

Section 1 introduces the theme of fallacies; Section 2 presents a short introduction to

captology and to the description of the connections we identified between fallacious

arguments and some of the techniques used in persuasive technologies; Sections 3

and 4 present, respectively, our survey and our empirical evaluation, and discuss the

results we obtained. Finally, our conclusions and a discussion of future work are

presented in Section 5.

2. Arguments in Logics

Logic1 is “the discipline studying the theory of valid inferences2. An inference is

composed by a set of initial propositions (premises) from which other propositions

(conclusions) are derived. All the valid rules of classical logic are based on deductive

inferential schemes where the conclusion C is a logical consequence of the set of

premises ‹P1, Pn›. An example of deductive inference is the following:

P1: All the men are mortal

1 Here with this term we refer to the classical formal logic. 2 For the sake of simplicity, we will consider here the term “inference” as a synonym of the term “argument” (Nolt et al., 1998).

Influencing the Others’ Minds

89

P2: Socrates is a man

C: Socrates is mortal

However, not all the inferences are deductive and, therefore, logically valid (Cohen,

Cohen and Nagel, 1993). There are, in fact, several types of inductive3 inferences

where the conclusion does not logically follow from the premises. An example of

inductive inference is shown below:

P: All the Mexicans that I know love Voltaire’s books

C: Mexicans love Voltaire’s books

Within the class of inductive inferences, logical fallacies enjoy a special status. In

fact, they are inferences that, “even if invalid from a formal point of view, appear as

plausible and therefore are psychologically persuasive” (Cohen et al., 1993; Hamblin,

1970). According to this definition, then, not all inductive inferences can be considered

as fallacious. An important aspect to point out regards the connection between

inferential validity and rationality: a fallacious argument, in fact, is not necessarily

“irrational”. Indeed, since the psychological/cognitive aspect plays a crucial role in the

dynamics of persuasion, a fallacious argument is usually an invalid argument endowed

with psychological plausibility and a proper heuristic value.

From an historical perspective, the study and classification of logical fallacies goes

back to the Philosopher in the De Sophistichis Elenchis (Aristotle, 1995). During the

centuries different research areas such as logic, rhetoric and argumentation theory

dealt with the problem of fallacies, pointing out that fallacious arguments are suitable

to be used as techniques for achieving persuasive goals (Perelman, Olbrechts-Tyteca

and Meyer, 1958). In addition, it is worth mentioning the attempts that, during the

centuries, different scholars have pursued in order to design “persuasive machineries”

or mechanisms able to influence the human audiences through the presentation of

particular combinations of logical and paralogical arguments. Examples of this case

are the Ars Magna of Ramon Llull (Bonner, 1985), that later widely influenced also the

works of Giordano Bruno and Gottfried Leibniz.

In the last 30 years, in the field of argumentation theory, a number of criticisms have

been raised about the use of classical logic as an instrument for the analysis of

3 For the sake of simplicity, here we will refer to all the inferences that are not deductive with the term “inductive inference”. Therefore even the abduction, in this case, can be ascribed to the category of “inductive inferences”.

A. Lieto and F.Vernero

90

fallacious arguments, and some alternative solutions have been proposed in order to

justify the use of such arguments in certain contexts (e.g. in the case of the “New

Dialectic” approach proposed by Douglas Walton (Walton, 1995). However such

criticisms have, in our opinion, some limits. More specifically: i) they do not allow to

characterize the difference between fallacies, errors, and weak arguments, and (ii) the

risk of “relativism” seems to be around the corner since these approaches hypothesize

contexts where the traditional fallacies are no more considered “fallacious”. For these

reasons, in the following, we present the link individuated between fallacious

arguments and persuasive technologies.

3. Fallacies and Persuasive Technologies

In the Nineteen Nineties, B.J. Fogg (Fogg, 2003) coined the term “captology” as an

acronym for the expression “Computers As Persuasive Technologies”, to describe a

research area which regards computer technologies as potential persuaders and

focuses on both their analysis and their design. According to Fogg, persuasion can be

defined as an attempt “to change attitudes or behaviors or both (without using coercion

or deception)” (Fogg, 2003). Following on from this definition, all computer

technologies which are purposely designed with the aim of changing their users’

attitudes or behaviors can be considered as persuasive (Fogg, 2003).

In the field of captology, the above mentioned connection between fallacies and

technology-based persuasion has been firstly pointed out in our previous work (Lieto

and Vernero, 2013), where we carried out a preliminary investigation, aimed at

recording the use of different fallacy-based persuasion strategies in existing e-

commerce websites. However, to the best of our knowledge, there is still lack of wider

empirical studies, performed on different technological environments, aimed to confirm

(or disconfirm) such an hypothesis.

The analysis on mobile apps and websites presented in this paper can be seen, then,

as an effort to provide a wider empirical framework to the assumptions that were

proposed in (Lieto and Vernero, 2013).

In the rest of this section we will present the connections that we identified between

some well-known logical fallacies and some of the techniques used in the field of

persuasive technologies.

Influencing the Others’ Minds

91

The logical fallacy known as argumentum ad populum, or “appeal to the majority”,

consists in accepting a certain thesis based on the mere fact that most people accept

it. A typical example of such a fallacy is: “Most people like a certain book, then that

book is worth-reading”.

This fallacy can be compared to those strategies, commonly used in the realm of

persuasive technologies, which owe their persuasive potential to the exploitation of

social dynamics. For example, technologies which grant access to social networks can

leverage influence dynamics among peers to stimulate their users to attain certain

goals. More specifically, Fogg refers to well-known social psychology theories, such as

social comparison and conformity (Turner, 1991), which can be applied to computer

technologies. According to social comparison theory, for example, people who are

uncertain about the way they should behave in a situation actively seek information

about others and use such information to form their own attitudes and behaviors.

Conformity theory, on the contrary, focuses on normative influence, claiming that

people who are part of a group usually experience a pressure to conform to the

expectations of the other members of their group.

A further commonality with fallacies can be found focusing on the discussion about

credibility that characterizes the area of persuasive technologies (Fogg, 2001). The

perceived credibility (and, therefore, persuasiveness) of both people and computers is

known to be affected by the so-called halo effect (Dion, Berscheid and Walster, 1972),

according to which a positive evaluation with respect to a certain feature (e.g., physical

attractiveness) produces a “halo” which causes an extension of such an evaluation to

other, unrelated, features (e.g., expertise).

Similarly, the fallacy of argumentum ad verecundiam (also “appeal to authority”)

refers to cases where some theses are assumed to hold based on the fact that the

person asserting them is wrongly assumed to be an authority about the topic of the

discourse because of his/her achievements in other, unrelated, fields. An example of

such a fallacious argument is the following: “the economist X claims that vegan diet is

dangerous for our health. Therefore: it is wrong to follow vegan diets”.

Technologies which implement tailoring techniques are persuasive because they

provide each individual with the information they are likely to find the most interesting,

based on their personal preferences, goals and experience. Obtaining personalized

information does not only save users the effort to examine an overwhelming amount of

content, but it is also more likely to draw their attention and, in case the so-obtained

information is accepted, it can determine deeper and longer-lasting changes. Various

A. Lieto and F.Vernero

92

personalization techniques are commonly adopted in adaptive systems and in

recommender systems, such as collaborative and content-based filtering (Adomavicius

and Tuzhilin, 2005). Personalization techniques can be considered fallacious because

they are based on the assumption that (i) people will maintain their past preferences in

the future (content-based filtering) or that (ii) people who have proved to have similar

preferences in the past will maintain this similarity also in the future (collaborative

filtering), which, although being probable, cannot be taken for granted.

Tailoring can be compared to the so-called audience agreement technique, which is

well known in rhetoric and theory of argumentation (Perelman et al., 1958). According

to this technique, persuaders should only use arguments which have already been

accepted by their audience in order to be effective.

Differently, according to the argumentum ad consequentiam, a proposition is

accepted based on the desirability or undesirability of its consequences (a positive

example of this fallacy is: “If there is an afterlife, then we will meet our loved ones

again. Therefore: there must be an afterlife”).

In the field of persuasive technologies, allowing users to explore cause-and-effect

relationships is a well-known technique, which exploits the possibility to offer computer

simulations where users can manipulate certain inputs (e.g., their daily food intake)

and observe their consequences (e.g., changes in their weight) (Fogg, 2003).

Prominent examples which show how cause-and-effect simulations can be used with

persuasive effects can be found in environmentalist websites which allow users to

calculate their ecological footprint (i.e., the number of planets which would be needed

if everyone lived like them) based on their lifestyle and consumption habits. Similarly,

in some online shops, users might be able to virtually try on a piece of clothing in order

to anticipate how they would look like if they bought it and wore it.

Finally, the accent fallacy, which occurs when emphasis is used to manipulate the

actual meaning of a proposition, is commonly adopted with a persuasive intent in

computer technologies, especially in its visual variant where certain elements are

made more visually prominent in order to emphasize them. A common example of the

(visual) accent fallacy occurs when special offers (e.g., discounts) are highlighted with

big fonts and bright colors, while the possibly restrictive conditions to enjoy them are

made scarcely visible. In Human-Computer Interaction, the accent fallacy can be

compared to misplaced salience, which is known as one of the “demons” hindering

situation awareness (Endsley, Bolt and Jones, 2003). While appropriate salience can

help to identify the most important information in a certain context, misplaced salience

Influencing the Others’ Minds

93

emphasizes irrelevant cues, confusing users and leading them to inappropriate

behaviors.

In our previous work (Lieto and Vernero, 2013), we had also pointed out some

similarity between surveillance technique and the argumentum ad baculum fallacy.

Surveillance is based on the idea that people tend to change the way they behave

when they are aware that they are being observed, especially if the observer has the

power to punish or reward them (in this case, they will tend to match the observer’s

expectations) (Turner, 1991). The covert menace which underlies surveillance

technique is not too dissimilar to the argumentum ad baculum, where the persuader

resorts to threats of force in order to make his/her thesis be accepted. An example of

this fallacy, inspired to Pascal’s Gamble (Pascal, 1864), is: “If you don’t believe that

God exists, when you die you will be judged and sent to Hell, so it is safer to believe in

God”. It is important to notice, however, that the use of some form of coercion is

borderline with respect to Fogg’s definition of persuasion (Fogg, 2003). Moreover, in

(Lieto and Vernero, 2013) we had observed that only a very small percentage of the

websites we had examined made use of persuasion strategies which could be mapped

to the argumentum ad baculum (e.g., making the actions performed on the website

totally “transparent”, so that users might be induced to buy products or services which

are consistent with the self-image they want to show to others). For these reasons, we

will not consider this fallacy in our current work.

Fallacy Websites and App features

Arg. ad populum Best seller products, ratings

Arg. ad verecundiam Improper testimonials

Audience agreement Personalization

Arg. ad baculum Public visibility of purchased/browsed items or wish lists

Arg. ad consequentiam Cause-effect simulations

Accent Emphasis/hiding of information

Table 1. Correspondence  matrix between fallacious arguments and websites/mobile-apps displayed features.

4. Survey on Persuasion Techniques in e-Commerce Apps and Websites

In order to investigate whether, among the techniques used in persuasive

technologies, there actually are some which are reducible to arguments based on

logical fallacies, as hypothesized in (Lieto and Vernero, 2013), we carried out a survey

A. Lieto and F.Vernero

94

on 101 mobile apps and 175 websites in the e-commerce domain4. In fact, we

surmised that technologies with a clear persuasive goal (i.e., selling goods) should

make an extensive use of persuasive techniques.

Figure 1. An example of the arg. ad populum fallacy in Comtech.de (a) and Edeka 24 (b).

Our set of websites, which extends the one examined in our previous work (Lieto and

Vernero, 2013), was collected by searching for “online shopping” on Google and

selecting all pertinent results. As far as apps are concerned, we first searched for apps

directly related to the e-commerce websites on the Apple App Store, and found 71 of

such apps. The remaining 30 apps were identified by querying the App Store for

“online shopping”, similarly to what we did for websites5.

In order to run the evaluation we created a correspondence table (see Table 1)

where the presence of fallacious arguments is connected to the use of some specific

features in the examined apps and websites.

As shown in Table 1, the argumentum ad populum has been associated to the case

in which either “best seller” products (see Figure 1) or user ratings are displayed (in

this case the persuasion strategy is based on the following argument: “Most people

buy/like X, then it is positive to buy X”). The argumentum ad verecundiam has been

associated to the presence, in one or more parts of the websites and apps, of improper

testimonials for certain products, and the audience agreement has been associated to

4 The complete list of websites and apps analyzed is available at: https://sites.google.com/site/techsuasion/. 5 Some preliminary results regarding our analysis of mobile apps were presented in (Lieto and Vernero, 2014).

Influencing the Others’ Minds

95

the use of recommendation techniques (see Figure 2). Finally, the argumentum ad

consequentiam has been associated to the presence of software environments which

are able to simulate the consequences of certain user choices, and the accent fallacy

to the case when part of the purchasing-related information is emphasized and part is

hidden (e.g. when shipping or tax costs are presented only at the end of the

purchasing process, or when certain products are given more visual prominence than

others).

Figure 2. An example of the audience agreement fallacy in AllSaints, (a, content-based

suggestion) and Edeka 24 (b, collaborative-filtering-based suggestions).

4.1 Results

Table 2 shows the obtained results. We have recorded no use of fallacious-reducible

arguments only on the 16% of mobile apps and on the 13,1% of websites in our set.

The most recorded fallacies result to be the accent (apps: 57%; websites: 54,3%), the

argumentum ad populum (apps: 37%; websites: 49,7%) and the audience agreement

(apps: 21%; websites: 36%). Notice, however, that the percentages of adoption of the

argumentum ad populum and the audience agreement fallacies drop significantly in

the case of mobile apps. Concerning the least observed fallacies, the argumentum ad

verecundiam appears more often in websites than in apps (where it obtains the lowest

count), and the same happens for the argumentum ad consequentiam. It should be

observed that, while the use of the other fallacies does not seem to be related to the

type of items sold by a certain website or application, the argumentum ad

A. Lieto and F.Vernero

96

consequentiam is used almost only by shops selling clothing, shoes and make-up

products, i.e., items for which consequences are relatively easy to anticipate and

simulate.

In order to better understand whether the differences we observed between websites

and apps should be ascribed to device-dependent peculiarities on the use of

fallacious-reducible techniques or to some bias due to more general differences

between the two sets, we focused on the 71 websites and on the 71 apps which are

directly related to each other (i.e., they were designed to promote the same

company/brand). Results for this second analysis, which are quite close, in general, to

those we obtained for the whole dataset, are presented in Table 3. Accent maintains

its primacy as the most popular fallacy among both websites and apps, while

persuasion strategies based on the argumentum ad populum and the audience

agreement fallacies still appear to be adopted less often in mobile apps than in the

corresponding e-commerce websites, thus confirming and actually expanding the

adoption gap we could observe in our whole dataset. This difference is especially clear

for the argumentum ad populum, which could be observed only in 28,2% of the apps,

while it was much more popular among websites (50,7%). On the contrary, the

adoption gap between apps and websites is slightly smaller in the reduced than in the

whole dataset if we consider persuasion strategies based on the argumentum ad

consequentiam. Functionality and features which implement this fallacy are probably

considered distinctive for a certain company/brand, so that, when adopted, they are

applied cross-platform.

Our data also suggest that apps are usually endowed with less fallacious and

persuasive features than websites, an insight which could be explained by the fact that

mobile applications are a more recent, not yet fully mature, technological environment.

Fallacy % – apps % – websites

Arg. ad populum 37% 49,7%

Arg. ad verecundiam 3% 13,1%

Audience agreement 21% 36%

Arg. ad consequentiam 4% 8,6%

Accent 57% 54,3%

None of the above 16% 13,1%

Table 2. Percentages of e-commerce apps and websites using fallacious-reducible persuasive mechanisms.

Influencing the Others’ Minds

97

Fallacy % – apps % – websites

Arg. ad populum 28,2% 50,7%

Arg. ad verecundiam 2,8% 15,5%

Audience agreement 25,4% 42,3%

Arg. ad consequentiam 5,6% 8,5%

Accent 56,3% 59,2%

None of the above 21,1% 7%

Table 3. Cross-platform comparison of 71 directly-related apps and websites.

5. Experimental Evaluation

Having observed that most of the examined e-commerce apps and websites make

use of at least one persuasion strategy based on fallacious arguments, we carried out

an experimental evaluation aimed at empirically assessing the relative effectiveness of

fallacy-based strategies in a controlled environment.

As a use case, we chose a fictional online bookshop. We concentrated on a website

(rather than a mobile app) for various reasons: on the one hand, websites require no

installation on the part of users and provide a familiar, uniform interaction modality

across different software/hardware platforms; on the other hand, people are still more

likely to do their shopping using a website than a smartphone app6. For simplicity of

implementation, and because they can be adopted independently of the kind of goods

that are sold, we chose to focus on four fallacies out of the five we examined in our

survey: accent, argumentum ad populum, argumentum ad verecundiam and audience

agreement.

Our evaluation was divided in two phases: first, we carried out a pilot study where

participants interacted with a simple website where no persuasion strategies were

adopted. This study was meant to highlight how users choose books to buy in a non-

persuasive context and served as a control condition. Then, we conducted a larger

study with a “persuasive” version of the same website implementing design elements

directly related to fallacious arguments as showed in the correspondence matrix

presented in Table 1.

6 According to the 2013 “Local Media Tracking Study” conducted by marketing research company Burke (http://www.burke.com/) in the U.S.A., consumers increasingly turn to mobile devices for their online shopping. However, about seven consumers out of ten prefer using mobile websites than apps.

A. Lieto and F.Vernero

98

5.1 Participants

Seven people participated in our pilot study and 19 in the larger one. All of them were

selected from the same population, i.e., students and colleagues at the Department of

Computer Science, University of Turin, using an availability sampling strategy. They

are 42% female and 58% male, aged 22-45. They read 8,5 books per year on average

and most of them (77%) have bought books online at least once. Their preferences

about literary genres are quite variegated, with 35% of the participants being

passionate about adventure fiction, 19% not loving it, but being keen on some

adventure-related genres (e.g., fantasy or science-fiction) and the remaining 46%

having different tastes (e.g., essays).

5.2 Material

Having decided to carry out our experiment in the context of an online bookshop, we

prepared some basic information to present ten imaginary books belonging to the same

genre, seafaring adventures: the title, the name of the author and a short description.

All information was made up and we paid attention that it followed a similar format for

all the books, in order to limit the number of factors which might influence participants’

choices. For example, we had a book entitled “Docking at Sibyl Island (Attracco

all’Isola della Sibilla)”, written by the imaginary author “S. Sand”, with the following

short description: “This book deals with a seafaring adventure on Sibyl Island”.

Moreover, all books had the same price (10.50 €).

Figure 3. A screenshot of the non-persuasive version of the online bookshop website.

Influencing the Others’ Minds

99

In the non-persuasive version of our online bookshop, organized according to a one-

column layout, books were simply listed in alphabetical order, depending on their title

(see Figure 3). Each book was accompanied by a “Buy” button, pressing which

participants landed to a confirmation page that recapped the details of their (simulated)

purchase.

In the persuasive version of our online bookshop, which had a similar look and feel to

the non-persuasive one, the main page presenting all available books was preceded by

an intro page where we explained that the bookshop offered personalized

recommendations and asked participants to choose their favorite book between two

classics, “The Pirates of Malaysia” by Emilio Salgari and “Moby Dick” by Herman

Melville, in order to provide the website with some information about their preferences

(both books were accompanied by a short summary of their plot, drawn from the

Internet). In the bookshop main page, which was organized according to a two-column

layout, four books were presented in a peculiar way, following the four fallacious

persuasion strategies we experimented with (see Figure 4):

• Accent fallacy: a book was presented on top of the page, in a box named “Our

highlight for this month” and extending across the two columns. It was given visual

prominence through the use of a larger font size and an accompanying badge with

the text “book of the month”.

• Argumentum ad populum fallacy: a book was presented in a box named “Our best

seller”, in the right-side column. It was accompanied by a sentence explaining that

it was the best-selling book among the customers of the bookshop.

• Audience agreement fallacy: a book was presented in a box named “Chosen for

you”, in the right-side column. A short sentence explained that it had been selected

according to the participant’s preferences.

• Argumentum ad verecundiam fallacy: a book appearing in the alphabetically-

ordered list in the left-side column was accompanied by the photo of a celebrity

and a comment of theirs, stating that it was “the best adventure book ever”. We

used a photo of Mario Draghi (an economist and the current President of the

European Central Bank) in half the cases and of Fernando Alonso (a Spanish

Formula One racing driver, racing for Scuderia Ferrari at the time of our

experiment) in the other half.

A. Lieto and F.Vernero

100

Figure 4. A screenshot of the persuasive version of the online bookshop website (fallacious persuasion strategies are highlighted).

The book to present as a personalized suggestion (audience agreement) was

selected according to the similarity of its title to that of the classic chosen by a certain

participant in the intro page, while the books connected to the other three fallacies were

chosen at random for each participant. The remaining six books were simply presented

in alphabetical order, in the left-side column.

5.3 Procedure

The study took place in a room at the University of Turin, one participant at a time.

Instructions were provided in written form, according to the following script: “Our

Bookshop is specialised in adventure novels. Imagine you have decided to buy yourself

a book of this genre for Christmas. Which one would you choose? Feel free to think

aloud while you are exploring the bookshop website. Mind: you can choose a single

book. Once you have made your choice, please select “buy” to conclude the

experiment. Don’t worry: this is just a simulation and you will not be charged any

amount.”

The experimenters silently observed participants’ interaction with the website,

intervening only in case they were posed some explicit question. Moreover, they noted

Influencing the Others’ Minds

101

down all the comments participants expressed during the evaluation. After participants

had chosen a book to buy, one of the experimenters carried out a short interview aimed

at collecting information about their demographics and reading-related habits.

5.4 Results

In the pilot study, the seven participants chose five different books, namely, they

made quite heterogeneous choices, as it can be expected when there are no other

factors than personal preferences which can exert their influence. Gini Heterogeneity

Index, a measure of dispersion for categorical variables which ranges from a minimum

of 0 to a maximum of 1, equals 0,86, a value indicating high dishomogeneity in

participants’ behavior. Moreover, in their thinking aloud, all participants actually

explained their choices based on the appeal of the title.

In the second study, involving the interaction with the persuasive (fallacy-based)

version of the website, a record of participants’ choices was collected. The results of

this analysis are presented in Figure 5: 47% of the participants chose a book presented

according to one of the four persuasion strategies, and motivated their choice

accordingly in their thinking aloud. Participants who chose the “book of the month”

(accent fallacy, 26%) explained either that it was the first one they had noticed, or that,

dealing all the books with similar topics, they trusted what seemed to be a

recommendation on the part of the bookshop owners. Participants who chose a book

which was recommended to them based on their previously-expressed preferences

(audience agreement fallacy, 16%) motivated their choice with the very fact that they

were confident to like such a personalized suggestion. Finally, the only participant who

chose a book promoted by a celebrity (in her case, Mario Draghi) was actually

impressed by his endorsement. The remaining participants, who chose books not

presented through any persuasion strategy (53%), motivated their choice with

arguments referring to the appeal of the title, as it happened in the pilot study.

Our results suggest that persuasion strategies based on logical fallacies actually have

an effect on people’s behavior, at least in the context of an e-commerce website. As we

have observed before, in fact, people only took into account intrinsic item properties (in

our case, the book title) in a situation where no persuasion strategies were used, while

they were guided by fallacious heuristics (i.e., they used different criteria to evaluate

the available options) in almost half of the cases in a persuasive environment.

A. Lieto and F.Vernero

102

Figure 5. Distribution of user choices with respect to the persuasion strategy used for book presentation.

Comparing these results with those of our survey (see Section 4), we noticed that the

effectiveness of persuasion strategies, according to our empirical evaluation, seems to

be proportional to their popularity in existing e-commerce websites and apps, with

accent being the top strategy in both cases, followed by the audience agreement and

argumentum ad verecundiam fallacies. Interestingly, however, we also noticed a big

exception: the argumentum ad populum fallacy, used by almost 50% websites and

about 37% apps, was totally ineffective in our evaluation. Not only books presented as

the “best sellers” were not chosen by any participant, but a couple of them also stated

that knowing what other people had bought was useless to them, and that suggestions

based on the preferences of other people were far less relevant than personalized

ones. This somehow unexpected fact might be explained considering that all our

participants belong to an academic environment and are probably more critical

towards “blockbuster” books than the average person. In addition, we surmise that, in

some cases, participants might have provided socially desirable answers, avoiding to

choose best seller books to maintain their self-image of smart, critical and independent

thinkers.

6. Conclusion and Future Work

In this paper we have presented the results of a double empirical analysis aimed at

investigating both the actual use of fallacious-reducible arguments in persuasive

Influencing the Others’ Minds

103

technologies (such as e-commerce websites and mobile applications) and their efficacy

in influencing human decision making when implemented, with an explicit persuasive

goal, in an e-commerce website.

Our results show that fallacy-based persuasion strategies are extensively used in

existing digital artifacts, with some subtle differences between websites and mobile

apps, and that – at least for the case of e-commerce websites – they are actually

effective in influencing users’ behavior, offering shortcut heuristic criteria to ease their

decision-making. In particular, the strategy based on salience manipulation (accent

fallacy) resulted to be both the most popular and the most effective one.

As a short-term future goal, we are planning to carry out an empirical evaluation (by

following the same rationale used for the evaluation of the influence of fallacy-based

strategies in the e-commerce websites) based on the observation of the behavior of

users browsing a persuasive and a non-persuasive version of the same mobile app.

Similarly, we are considering to re-run our evaluation with a less homogeneous set of

participants in terms of their educational background: this would allow us to determine

whether the unexpected ineffectiveness of the persuasion strategy based on the

argumentum ad populum, which we observed in our present study, should be ascribed

to the peculiar features of our participants or to a more general trend.

In addition, we plan to extend our analysis by increasing the number of both the

examined logical fallacies (and the correspondence matrix between fallacies and

persuasion techniques we individuated), as well as the technological environments

where they have been (or can be) used. In particular: the social networking

environments and the systems based on avatar technologies adopting forms of

linguistic interaction seem to be ideal candidates for investigating the feasibility of our

approach in other technological scenarios.

From a more general perspective, the research carried out and resulting from this

work can be productive in at least two orthogonal lines of investigation. On the one

hand, in fact, it can be useful for the implementation of algorithms that – starting from

the individuated correspondence matrix between fallacies and displayed technological

features – can be used for the automatic or semi-automatic detection and classification

of fallacious-reducible strategies adopted on a large volume of websites, mobile

applications and other technologies. Such detection could be useful for individuating

cases of unethical use of fallacy-based strategies w.r.t. the end-users.

On the other hand, the results obtained with the evaluation of the efficacy of fallacy-

based persuasion mechanisms in different technological contexts could be useful in

A. Lieto and F.Vernero

104

order to individuate standard “persuasions patterns”, and the most efficacious

combinations of them, that can be directly used as a basis for the design of novel

persuasive technologies. Since the application of such patterns should be, as

mentioned, delegated only to ethically acceptable uses of persuasive technologies, the

two lines of research individuated above, despite different, will require a continuous

interaction.

Finally, in case we identify logical fallacies not yet finding a correspondence with

existing persuasion strategies used in the field of captology, these could serve as a

basis for the design of new persuasive features and patterns.

7. References

Adomavicius, G., & Tuzhilin, A. (2005). Toward the Next Generation of Recommender

Systems: A Survey of the State-of-the-Art and Possible Extensions. IEEE

Transactions on Knowledge and Data Engineering, 17(6), 734-749.

Aristotle, J.B. (1995). Complete Works of Aristotle. Princeton: J. Barnes.

Bonner, A. (Ed.) (1985). Selected Works of Ramon Llull (1232-1316). Princeton:

Princeton University Press.

Cohen, M.R., Cohen, M.R., & Nagel, E. (1993). An Introduction to Logic. Cambridge,

MA: Hackett Publishing.

Dion, K., Berscheid, E., & Walster, E. (1972). What is Beautiful is Good. Journal of

Personality and Social Psychology, 24(3), 285-290.

Endsley, M.R., Bolt, B., & Jones, D.G. (2003). Designing for Situation Awareness: An

Approach to User-Centered Design. London and New York: Taylor & Francis.

Fogg, B.J. (2003). Persuasive Technology: Using Computers to Change what we Think

and Do. Amsterdam: Morgan Kaufmann.

Fogg, B.J., Marshall, J., Laraki, O., Osipovich, A., Varma, C., Fang, N., Paul, J.,

Rangnekar, A., Shon, J., Swani, P., & Treinen, M. (2001). What Makes Web Sites

Credible? A Report on a Large Quantitative Study. In Proceedings of the SIGCHI

Conference on Human Factors in Computing Systems (pp. 61-68). New York: ACM.

Hamblin, C.L. (1970). Fallacies. London: Methuen.

Lieto, A., & Vernero, F. (2013). Unveiling the Link between Logical Fallacies and Web

Persuasion. In Proceedings of the 5th Annual ACM Web Science Conference,

WebSci ’13 (pp. 473-478). New York: ACM.

Influencing the Others’ Minds

105

Lieto, A., & Vernero, F. (2014). Persuasion in Mobile Apps and Logical Fallacies: A

Preliminary Study. In Adjunct Proceedings of the 9th International Conference on

Persuasive Technology (pp. 40-42). Padova, Italy: CLEUP.

Nolt, J., Rohatyn, J., & Varzi, A. (1998). Introduction to Logic. New York: McGraw Hill.

Pascal, B. (1864). Oeuvres Complètes de Blaise Pascal. New York: Hachette.

Perelman, C., Olbrechts-Tyteca, L., & Meyer, M. (1958). Traité de l’Argumentation: La

Nouvelle Rhétorique, vol. 1. Paris: Presses universitaires de France.

Turner, J.C. (1991). Social Influence. Milton Keynes, UK: Open University Press,.

Walton, D.N. (1995). A Pragmatic Theory of Fallacy. Tuscaloosa: University of

Alabama Press.

Copyright of PsychNology Journal is the property of Luciano Gamberini and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder’s express written permission. However, users may print, download, or email articles for individual use.


Comments are closed.