=Paper=
{{Paper
|id=Vol-2885/paper8.pdf
|storemode=property
|title=Project-Machete: a Weapon to Cut Through the Amazon(.ca)
|pdfUrl=https://ceur-ws.org/Vol-2885/paper8.pdf
|volume=Vol-2885
|authors=F. Bennett W. Lewis
|dblpUrl=https://dblp.org/rec/conf/persuasive/Lewis21
}}
==Project-Machete: a Weapon to Cut Through the Amazon(.ca)==
https://ceur-ws.org/Vol-2885/paper8.pdf
Project-Machete: a Weapon to Cut Through the
Amazon(.ca)
F. Bennett W. Lewis[0000−0001−6929−4455]
University of Saskatchewan, Saskatoon SK, Canada
fbl773@usask.ca
Abstract. As persuasive systems seep into our lives on the back of technology,
the need for counter-persuasive technologies becomes more apparent. Many
users are not aware of design elements that exploit their cognitive biases and
encourage them to act against their own self-interests; Project-Machete was
created to level the playing field. It seeks to limit the ability of persuasive
technologies on the Amazon shopping platform to persuade users into making
impulsive or extrinsically influenced purchases. It removes biased plays to item
scarcity, selectively reveals recommended items, counters biased encouragement
to buy new over used items, and encourages shoppers to think critically about
whether a purchase should be made at all. Project-Machete achieves this through
JavaScript driven DOM-tree manipulation inside an extension for the Google
Chrome web browser.
Keywords: E-tail, Counter-persuasive Technology, Persuasive Systems Design,
Online Shopping, Amazon Online Shopping, Cognitive Bias, Behavior Change
Support Systems
1 Introduction
Ethics in Persuasive Technology (PT) is a largely fragmented, standardless, and treated
as an afterthought. In the analysis of a literary review conducted in 2019 regarding the
past 13 years of literature in PT, it was found that only 66/375 of the analyzed papers
on PT mentioned more than 1 sentence regarding ethics. Of those 66 papers, 43 of them
mentioned ethics “In passing” without discussion [5]. There are currently a handful of
proposals regarding ethics in PT, but nothing agreed upon as a standard [8,3] and while
HCI frameworks such as those proposed by Fogg [2] and Oinas-Kukkonen [6] exclude
deception and coercion by definition, this does not limit the use of Persuasive Design
Elements (PDEs) to exclusively noble ends. This paper argues under the supposition
that Amazon’s intent is not simply for it’s users to buy products, but to buy lots and buy
fast, and employs arguably deceptive and coercive technology achieve this. While
Amazon could claim that they are justified in their use of PDEs to provide a better
shopping experience to users, they can not deny the involuntary exposure to PDEs that
encourage the achievement of this goal. When Amazon’s PDE’s and a lack of user
Copyright © 2021 for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
Ninth International Workshop on Behavior Change Support Systems (BCSS 2021): 89
Project-Machete: a Weapon to Cut Through the Amazon(.ca)
awareness thereof meet, the powers of persuasive technologies are both powerful and
potentially exploitative. This is what Project-Machete seeks to address.
While we humans have had countless years experience with face-to-face encounters
with persuasion, deception, and coercion, we have not yet had a century of face-to-
screen encounters. Consider somebody in the midst of buying from a salesperson in a
brick-and-mortar shop. If this salesperson was constantly muttering phrases under their
breath, urgently trying to steer you towards or away from certain products, or
relentlessly trying to up-sell you, you would be naturally suspicious of them! This is
where Project-Machete (PM) comes in, to aid those who are not yet as suspicious of PT
in a sales setting. PM’s implementation and concept were inspired by Benjamin
Grosser’s “Facebook Demetricator” chrome extension [4]. Whereas Grosser’s
extension reduces the ability of users to measure their social worth through quantitative
social artifacts, PM reduces Amazon’s ability to persuade users into making purchases.
It filters out the noise and in doing so, acts as a “Digital-Conscience” for it’s users. This
is not to say that users do not have a digital conscience; that they are incapable of
making informed, critically-evaluated decisions online, but rather that they are not
aware of the mechanisms in place that shape these decisions; that their digital
conscience is underdeveloped. PM was created to aid this conscience, and to do so
ethically. To achieve it’s ethical goals, PM will seek to follow the guidelines proposed
by Spahn [8] in future development. Further details on ethical considerations will be
addressed later in Ethics & Future Work.
2 Implementation
As a content-script style Chrome extension, PM is allowed to inject a JavaScript (JS)
and a CSS file into a set of pre-defined domains. This is what allows PM to modify the
raw HTML of Amazon’s pages. At a high level, PM’s depersusification process goes
through 3 stages 1) Page Identification, 2) Item Identification, and 3) Item Mitigation.
Page identification matches specific pages on the Amazon domain using regex, once a
page is recognized stage 2 is entered. This stage uses JS and Amazon’s implementation
of best practice web-development (classes for common style elements, ids for special
items) to collect various PDEs on a page (i.e. recommended lists, the ”buy-now” button)
for stage 3. In stage 3, PM addresses PDEs as an item to either remove or modify.
Notable PT elements that are removed include the ”Buy now” button and the ”In stock”
report. The “Buy now” button is an example of reduction [2] while the “In stock” report
is an example of Cialdini’s principle of scarcity [1]. What should be noted is that this
report is deceiving with its content. This is because the number reported as “in stock”
neglects to take into account the availability of used items and often choose a niche-
option for a given item that is actually low on stock. Targets to be modified are
addressed by changing the behavior/appearance of Amazon’s HTML. While shopping
on Amazon.ca, one might notice that the prices for new items are displayed with larger,
bolder font, and are placed higher on an item posting than their smaller, hidden, used
counterparts (Figure 1). This seeming preference of Amazon that we buy new is
countered by PM, which auto-selects the “Buy used” option when it is available for an
90 Ninth International Workshop on Behavior Change Support Systems (BCSS 2021):
Project-Machete: a Weapon to Cut Through the Amazon(.ca)
item (Figure 2). PM also chooses to modify recommended lists as removing them could
critically reduce the users shopping experience and infringes on user autonomy (Figure
3). The final line of defence that PM offers to counter Amazon’s PDEs is a button-cover
that is placed over the “Place order” button (Figure 4). When clicked, the user is met
with a quiz regarding the appropriateness of the purchase; It asks 1) Should you make
this purchase? 2) Do you need to make this purchase now? 3) Can you afford to spend
$X on this purchase? And finally, PM asks for 3 reasons that justify the purchase. It is
anticipated that some rushed users may attempt to “buzz through” the quiz without
giving real answers. This is why the quiz requires that all fields are filled and delays
the user for one minute before allowing them to proceed. This should discourage a user
from filling out the form with garbage answers and encourage them to take time and
think critically about their purchase. The quiz also provides the user with a quick link
to get away from the page if they are worried that their will power will fail. It should
be noted that this form collects no user data. It’s only purpose is to induce critical
thought in the user.
3 Status and Future Work
Though Project-Machete is incomplete, it is functional and version 1.2 is currently
deployed to the Chrome Web Store. As user feedback is gathered and more PDEs are
identified, the list of elements to address in later versions of PM grows. A detailed list
of technical issues for Project Machete can be found in the repository’s issues on gitlab.
Chief among those are: Current reliance on RegEx, inability to buy kindle e-books, and
a small UI issue introduced by inserting the HTML to collapse recommend lists on a
page.
3.1 Ethics & Future Work.
Though it was not initially considered as such, by engaging in counter-persuasion, PM
itself takes on elements of PT. With this in mind, PM could be seen as a piece of PT
that dabbles in coercion and deception. It cannot be denied that PM infringes on user
agency. It removes some elements, decides buying used is preferable to buying new, it
hides elements by default, and forces a user to fill out a form before making a purchase
without first asking if this is their desire. What could be argued though is that because
Amazon infringed on user autonomy first through selective reporting, relentless up-
sells, etc. - PM is simply leveling the playing field. For example, the ”in stock” report
is deceptive in that it fails to report used items, and chooses whichever niche option is
of lowest stock to report on; is PM being deceptive when it removes Amazon’s
deception? Should counter-persuasive technology be considered coercive/deceptive if
it removes such elements from other platforms? This is a question out of the scope of
PM, all the same, PM recognizes that infringing on user agency and engaging in
Ninth International Workshop on Behavior Change Support Systems (BCSS 2021): 91
Project-Machete: a Weapon to Cut Through the Amazon(.ca)
Fig.1. This figure illustrates the before/after of search results for a specific book. Notice the
placement and stylistic elements of the the price for new items when compared to used. This is
an exploitation of cognitive bias. Also notice PM’s removal of the ”in stock” report; inspection
of the item reveals more than 30 used items as available where the report claimed there were only
10. This is a play to scarcity[1].
92 Ninth International Workshop on Behavior Change Support Systems (BCSS 2021):
Project-Machete: a Weapon to Cut Through the Amazon(.ca)
Fig.2. This figure illustrates the before/after of PM applied to the item page, which contains many
different PDEs. PM’s targets include: The “Buy Now” button, which is an example of
reduction[2], the “In Stock” report (an example of scarcity[1]), the social-media easy share links,
and the “free-shipping” countdown clock. Aside from the share links, each of these examples are
aimed at persuading the user to buy new, and buy fast.
Fig.3. This figure illustrates the before/after of PM applied to recommended lists. Amazon.ca
puts forth two different types of recommended lists. Some of which load dynamically on the
page, others are page specific. PM collapses both by default but only one can be collapsed such
that the original title of the list is preserved (i.e. “Frequently bought together”, “Top Cyber
Monday Deals”). The other is collapsed into a section titled “show more/less”. Recommended
lists appear on nearly every amazon page in their dynamically-loaded form and appear only once
one has scrolled to the bottom of the page. This is part of the reason that PM is called on a timer.
Ninth International Workshop on Behavior Change Support Systems (BCSS 2021): 93
Project-Machete: a Weapon to Cut Through the Amazon(.ca)
Fig.4. This figure illustrates the elements added by PM on the ”Place order” page, including the
digital conscience form. While PM does not stop a user from making purchases; it does make an
effort to slow them down and encourage purchase evaluation. To best achieve this a “button
cover” is placed over the purchase button that once clicked, prompts the user with a quiz. It is
anticipated that users may try to bypass the form by rapidly filling it out so not only is the user
forced to enter all fields, but they are delayed a full minute before the form can be dismissed. The
form also offers a means to quickly navigate away from the page. Data from the form is not
collected and there are no plans to do so in the future.
deception/coercion is a poor way to promote ethics in PT and counter PT alike. Future
work on PM will seek to bolster user agency, & bring attention to it’s goal: “To provide
a shopping experience on Amazon that is free of influence from PDE’s implemented
by parties with conflicts of interest in the buying process”. By making users aware of
this goal & how it is achieved, their acceptance of PM would imply recognition and
voluntary exposure to the techniques PM employs. Looking at PM this way, its coercion
can be seen as similar to the example provided by Smids regarding the seat belt [7],
though rather than the end goal being that users abide by PM’s “digital conscience
form” because it will not go away, they do so because they believe it is providing them
something valuable. This user recognition of the quality of PM’s introduced practices
stems from a desire to take a virtue-ethic approach in technological ethics [9]. The
practices introduced by PM are then not only to create a shopping experience with less
PT, but to interrupt and dissolve any externally-induced attitudes that a good shopping
experience is one where you buy new, fast, and more.
A primary concern regarding PM’s aim is that users may end up with the idea that
justifying a purchase on moral fronts is as simple as filling out a form and waiting a
minute. PM does not currently, and has no plans in the future of evaluating the
worthiness user-provided purchase justifications as this would open the door to even
more ethical concerns. However by not doing so, users may use whatever reasons to
“justify” their purchases while still getting PM’s “approval”. This is where the problem
94 Ninth International Workshop on Behavior Change Support Systems (BCSS 2021):
Project-Machete: a Weapon to Cut Through the Amazon(.ca)
is, if PM’s form is considered a mere ”moral approval stamp” by users, they may end
up seeing any answer they provide as justifying, and as a result buy more because of
this. This undesirable potential side effect is why user education should be at the
forefront of PM’s future work. A user that understands the purpose and value of the
form will be critical of their own provided reasons.
Taking into account PM’s goals of increased user agency (beyond enabling/
disabling the extension) and education, PM favours the guidelines proposed by Spahn
regarding ethics in PT [8]. These guidelines include: ”G1) Persuasion should be
focused on (real or counterfactual) consent. G2) Ideally, the aim of persuasion should
be to end the persuasion . G3) Persuasion should grant as much autonomy as possible”
[8]. Currently, PM loosely achieves G1 & G3 by giving the user the ability to control if
elements are shown, but fails to do so completely as some elements are removed without
user consent. A fully-realized Project-Machete would seek first to achieve the G3 which
would involve allowing the user to determine to what extent PDEs are mitigated by PM,
inadvertently achieving G1 as well. Additionally, if G1 & G3 are achieved for PM, they
too are achieved for Amazon. This is because PM introduces user consent and autonomy
through its varied level of PDE management to the Amazon platform. Spahn’s second
guideline is currently neglected by PM and must be developed against in the future. It
would not be right to encourage dependence on a digital conscience to discern what is
and is not PT. As Spahn proposed, PT should aim to educate its users and remove their
dependence, rather than foster it [8]. Future work on PM should focus on not just
mitigating PDEs, but on increasing user awareness & knowledge of them. PM should
not simply aid the users digital conscience, but develop it. If PM is allowed to be used
as a crutch, it will foster dependence within its users in discerning PDEs. This is why
G2 is so important. PM must take an educational approach that shows users how to
recognize PDEs on their own, allowing PMs value to extend to any website. What
perhaps most crucial to PM’s ability educational goal is that it eventually encourages
its own removal. PM will not be able to develop a digital conscience in its users if it
provides them one indefinitely.
Acknowledgements: I wish to thank Dr.Vassileva for encouraging me to submit to this
conference. I would also like to thank anonymous reviewers for their feedback and
suggested resources. Finally, I would like to thank Benjamin Grosser for his Facebook
demetricator. The idea for Project-Machete would not have been formed without his
groundwork.
References
1. Cialdini, R.B.: Influence: Science and practice, vol. 4. Pearson education Boston, MA (2009)
2. Fogg, B.: A behavior model for persuasive design. In: Proceedings of the 4th International
Conference on Persuasive Technology. Persuasive ’09, Association for Computing
Machinery, New York, NY, USA (2009). https://doi.org/10.1145/1541948.1541999,
https://doi.org/10.1145/1541948. 1541999
Ninth International Workshop on Behavior Change Support Systems (BCSS 2021): 95
Project-Machete: a Weapon to Cut Through the Amazon(.ca)
3. Gram-Hansen, S.B., Gram-Hansen, L.B.: On the role of ethics in persuasive design. The
possibilities of ethical ICT p. 85 (2013)
4. Grosser, B.: What do metrics want? how quantification prescribes social interaction on
facebook. Computational Culture 4 (2014), http://computationalculture.net/ what-do-metrics-
want/
5. Kight, R., Gram-Hansen, S.B.: Do ethics matter in persuasive technology? In: International
Conference on Persuasive Technology. pp. 143–155. Springer (2019)
6. Oinas-Kukkonen, H.: A foundation for the study of behavior change support systems.
Personal and ubiquitous computing 17(6), 1223–1235 (2013)
7. Smids, J.: The voluntariness of persuasive technology. In: International Conference on
Persuasive Technology. pp. 123–132. Springer (2012)
8. Spahn, A.: And lead us (not) into persuasion...? persuasive technology and the ethics of
communication. Science and Engineering Ethics 18(4), 633–650 (2012).
https://doi.org/10.1007/s11948-011-9278-y
9. Verbeek, P.P.: Persuasive technology and moral responsibility toward an ethical framework
for persuasive technologies. Persuasive 6, 1–15 (2006)