PreprintPDF Available

Designing Trustworthy User Interfaces for the Voluntary Carbon Market: A Randomized Online Experiment

Authors:
Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

The voluntary carbon market is an important building block in the fight against climate change. However, it is not trivial for consumers to verify whether carbon offset projects deliver what they promise. While technical solutions for measuring their impact are emerging, there is a lack of understanding of how to translate this data into interface designs that mediate the establishment of trust. With interaction between users and offset projects mainly happening online, it is critical to meet this design challenge. To this end, we designed and evaluated interfaces with varying trust cues for carbon offset projects in a randomized online experiment (n=244). Our results show that content design, particularly financial and forest-related quantitative data presented at the right detail level, increases the perceived trustworthiness, while images have no significant effect. We contribute the first specific guidance for interface designers for carbon offsets and discuss implications for interaction design.
Content may be subject to copyright.
Designing Trustworthy User Interfaces for the Voluntary Carbon
Market: A Randomized Online Experiment
Klaudia Guzij
Center for Digital Technology and
Management, Germany
klaudia.guzij@cdtm.de
Michael Fröhlich
Center for Digital Technology and
Management, Germany
froehlich@cdtm.de
Florian Fincke
Pina Technologies GmbH, Germany
orian.ncke@pina.earth
Albrecht Schmidt
Ludwig Maximilian University,
Germany
albrecht.schmidt@i.lmu.de
Florian Alt
Bundeswehr University Munich,
Germany
orian.alt@unibw.de
ABSTRACT
The voluntary carbon market is an important building block in
the ght against climate change. However, it is not trivial for con-
sumers to verify whether carbon oset projects deliver what they
promise. While technical solutions for measuring their impact are
emerging, there is a lack of understanding of how to translate car-
bon oset data into trustworthy interface designs. With interaction
between users and oset projects mainly happening online, it is
critical to meet this design challenge. To this end, we designed
and evaluated interfaces with varying trust cues for carbon oset
projects in a randomized online experiment (n=244). Our results
show that content design, particularly nancial and forest-related
quantitative data presented at the right detail level, increases the
perceived trustworthiness, while images have no signicant eect.
We contribute the rst specic guidance for interface designers for
carbon osets and discuss implications for interaction design.
CCS CONCEPTS
Human-centered computing !Empirical studies in HCI
;
General and reference !Experimentation.
KEYWORDS
trust, trustworthy interfaces, carbon markets, charitable giving, hci,
experiment
ACM Reference Format:
Klaudia Guzij, Michael Fröhlich, Florian Fincke, Albrecht Schmidt, and Flo-
rian Alt. 2022. Designing Trustworthy User Interfaces for the Voluntary
Carbon Market: A Randomized Online Experiment. In Designing Interactive
Systems Conference 2022 (DIS ’22), June 13-June 17, 2022, Virtual Event, USA.
ACM, New York, NY, USA, 14 pages. https://doi.org/10.1145/XXXXXXX.
XXXXXXX
Also with Ludwig Maximilian University.
Also with Ludwig Maximilian University, Bundeswehr University Munich.
Permission to make digital or hard copies of part or all of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for prot or commercial advantage and that copies bear this notice and the full citation
on the rst page. Copyrights for third-party components of this work must be honored.
For all other uses, contact the owner/author(s).
To appear in DIS ’22, June 13-June 17, 2021, Virtual Event, USA
©2022 Copyright held by the owner/author(s).
ACM ISBN 978-1-4503-XXXX-X/18/06.
https://doi.org/10.1145/XXXXXXX.XXXXXXX
1 INTRODUCTION
Carbon osetting is becoming increasingly popular for reducing
greenhouse-gas emissions and limiting the climate catastrophe
[
40
,
56
]. The voluntary carbon market – individuals or companies
osetting emissions without regulations requiring them to – can
become an important building block in the ght against climate
change [
5
]. For individuals to engage with carbon oset projects it
is important that they understand and trust the projects they invest
in. With carbon oset projects spread across the globe and the long
time period over which the impact takes eect, it is not trivial to
verify whether projects deliver what they promise. With public
demands for more transparency [
10
], there is considerable ongoing
eort to develop data-driven systems that allow the tracking of
projects in a transparent an veriable way (e.g. [
37
]). However,
one aspect neglected so far is how to present the vast variety of
possible information pieces about carbon oset projects so users
can evaluate their trustworthiness. This is particularly relevant for
interaction design research as online interfaces are the primary way
of how information on carbon oset projects is communicated. As
oset buyers cannot check the status of the projects themselves the
establishment of trust needs to be mediated through user interfaces.
The increasing demand for transparency and upcoming new ways
to measure data on carbon oset projects result in a need to explore
how to design trustworthy interfaces for them.
HCI research has identied ease of use, content design, and
visual design as important aspects to inuence trustworthiness of
user interfaces [
50
,
64
]. With no literature addressing interfaces
of carbon oset projects, the eld of charitable giving oers a
starting point. Similar to osetting, donors give money voluntarily
and it is dicult to verify the impact of any particular donation.
Charities can improve trust by increase transparency regarding
both nancial [
91
] and non-nancial data [
6
,
73
]. Recent work
shows that interface design can help establish a trust relationship
between donors and charities [
50
]. In particular, content design has
a strong inuence on the perceived trustworthiness of charitable
actions in online contexts [
50
,
82
]. As content design is domain-
specic, insights cannot be transferred to carbon osetting directly.
Thus, while we can assume a similar relevance of content design
to build trustworthy interfaces, we do not yet know which data
To appear in DIS ’22, June 13-June 17, 2021, Virtual Event, USA Guzij et al.
can serve as reliable trust cues for user interfaces of carbon oset
projects.
To address this gap, we set out to explore interface content
design at the example of forest-based carbon oset projects. We
designed website interfaces combining ndings from HCI research,
a qualitative pre-study (n=13), and sampling of existing websites
of established carbon oset sellers (n=7). With these interfaces, we
tested the eect of dierent types of data (nancial, images, forest)
and their level of detail (low, high) on perceived trustworthiness in
a randomized interactive online experiment (n=244). Participants of
the experiment were asked to evaluate the trustworthiness of four
forest carbon oset projects and select one project to invest in. The
interfaces for each project diered concerning the type of data and
the level of detail they provided. Our results show that not all data
types are equally eective as trust cues. Quantitative forest data
– such as number of planted trees and captured carbon – stands
out as the strongest predictor of perceived trustworthiness and
project selection. Our results also show the importance of assessing
the level of detail for each data type separately. For example, at a
high level of detail the interface showing nance-related data was
chosen signicantly more often and received signicantly higher
ratings of trustworthiness than at a low level of detail. Providing
images of the tree planting progress did not aect project choice or
perceived trustworthiness of the project.
Our results are in line with research from the eld of charitable
giving, where quantitative data was shown to improve perceived
trustworthiness of charitable projects. From our results we derive
and discuss theoretical and practical implications for researchers
and interface designers. Doing so, this paper is the rst to report
specic guidance for interface designers on what data may serve
as trust cues in the domain of carbon osets. More generally, our
ndings shed new light on the question of how designers can create
online experiences that facilitate trust through interfaces in situ-
ations where users cannot check the fulllment of a transaction
directly. The growing interest in the voluntary carbon market along
with the public demand for more transparency will increase the
need to improve interfaces of carbon oset projects. Our research
points to the importance of focusing HCI research on trust-building
user interfaces that guide users through complex data with appro-
priate design. Future work may build on the ndings to develop
guidelines on how to transparently communicate the progress of
carbon oset projects.
Contribution Statement:
The contribution of this work is threefold. (1) It explores dif-
ferent types of trust cues for carbon oset projects and presents
six website interfaces. (2) Using these interfaces it evaluates the
inuences of these trust cues, specically the type of data (nancial,
images, forest) and their level of detail (low, high), in a randomized
online experiment (n=244). (3) Finding signicant dierences re-
lated to project choice and perceived trustworthiness, it discusses
theoretical and practical implications and is the rst work to report
design guidance related to building trustworthy interfaces in the
domain of carbon osets.
2 BACKGROUND & RELATED WORK
Our work draws from several strands of research, most notably
from the elds of charitable giving and trustful interfaces.
2.1 The Voluntary Forest Carbon Market
Carbon osetting refers to the process of funding the avoidance
of emissions through climate protection projects and thus com-
pensating for carbon emissions that arise from industrial or other
human activities [
10
,
44
]. Carbon oset projects are increasingly
demanded in the so-called voluntary carbon market, where private
organizations, communities, and individuals buy credits with the
aim to reduce their carbon footprint to claim „carbon neutrality“
without regulation demanding them to. Oset sellers can either
be private or governmental organizations that manage administra-
tive processes of a climate protection project. [
10
,
68
]. While there
are several types of carbon oset projects this paper focuses on
forest-based ones. Forest-based projects account for the highest
share in nature-based carbon oset projects, in which osetting
is performed through avoided deforestation (emission avoidance)
or aorestation (emission sequestration) [
66
]. Despite their po-
tential and impact, academic literature and press coverage have
documented various challenges and critical problems of forest car-
bon osets, especially in the context of project selection and data
communication [
54
,
55
,
78
]. Particularly, a lack of transparent data
has been a topic of current reports as oset buyers cannot easily
access data in order to verify a project’s eectiveness [10, 23, 37].
While there are ongoing eorts to meet these challenges and
establish more transparent [] and veriable [] methods to measure
the impact of carbon oset projects, we were not able to identify
research looking at the human side – i.e. what information needs to
be provided by interfaces so people can evaluate the trustworthiness
of carbon oset projects. With a public demand for more transparent
communication, this makes forest-based projects an interesting
example at which explore the design of interfaces that facilitate the
establishment of trust.
2.2 Trustful User Interfaces
Trust is a broadly used concept. In the following we provide a
denition for the context of this work and summarize existing
research on trustworthy user interfaces.
2.2.1 Definition of Trust and Trustworthiness. Trust is a multi-
faceted concept, combining emotional, cognitive, and behavioral
dimensions [
12
,
89
]. Cho et al. studied the multi-faceted meanings
of trust and dene it as follows [13] :
“Trust is the willingness of the trustor (evaluator) to take risk based
on a subjective belief that a trustee (evaluatee) will exhibit reliable
behavior to maximize the trustor’s interest under uncertainty (e.g.
ambiguity due to conicting evidence and/or ignorance caused by
complete lack of evidence) of a given situation based on the cognitive
assessment of past experience with the trustee.” (p. 28:5)
Sheppard & Sherman further identify key aspects of why and
when trust is needed and under which circumstances it is formed,
based on various multidisciplinary contexts [
77
]. Generally, trust is
Designing Trustworthy User Interfaces for the Voluntary Carbon Market To appear in DIS ’22, June 13-June 17, 2021, Virtual Event, USA
created from vulnerability that is needed in the presence of risk and
interdependence. [
51
,
70
,
77
]. Trustors (the ones trusting) will only
accept vulnerability if they have strong expectations that a certain
outcome will be achieved and that the trustee (the one trusted in)
does not exploit their vulnerability [
2
,
30
]. Sociological research, as
well as research in computation and automation, state that the need
for trust is based on the goal to reduce complexity and uncertainty
[41, 53].
According to various researchers, trustees have to fulll specic
trustworthiness beliefs for trust to be established. Being antecedents
to trust, the three most relevant trustworthiness beliefs are com-
petence,benevolence, and integrity [
31
,
32
,
57
,
62
,
71
,
74
,
84
,
85
].
Competence describes the ability to perform an action the trustor
needs; benevolence means that the trustee acts in the trustor’s best
interest; integrity deals with the trustee’s honesty and promise-
keeping [
58
]. This paper adopts a multidimensional denition of
trust and thus sees the three beliefs as antecedents that need to bet
met in order to establish trust.
2.2.2 Trustworthy User Interfaces. Trust and trustworthy interfaces
have been a topic of interest in interaction design research and at
DIS (see e.g. [
17
,
18
,
42
,
90
]). Online contexts require trustors to rely
on their trustee’s computer-mediated communication (their web-
site), which ultimately increases the complexity of trust [
1
,
9
,
43
,
75
].
Whereas the previously mentioned antecedents of trust represent
cognitive factors that can be applied more generally, the context of
online interfaces adds a layer of functional factors that inuence
trustworthiness. The main factors that inuence trustworthiness
in online contexts are ease of use,content design, and visual design
[
84
]. Ease of use examines the eort that a user feels when using an
(online) system [
19
,
20
]. Trustors who nd the navigation of a web-
site easy perceive it as more trustworthy [
82
]. However, research
also found that there is a subjective factor to perceived ease of use
[
38
]. Visual design refers to the aesthetic appeal of a website [
41
].
Content design focuses on creating trustworthiness by providing
information that is perceived as useful and comprehensive [
76
].
Trustworthiness of user interfaces increases when they provide
both comprehensive information and convey expertise and honesty,
without creating bias [25, 53].
Scholars studied the interplay of these three antecedents of trust-
worthiness [
64
,
82
,
83
]. Pengnate and Sarathy drew from Norman’s
emotional design framework and found that visual appeal is impor-
tant to create initial trust signals for a website’s rst impression
[
61
,
64
]. Thielsch et al. go beyond rst impressions and study de-
terminants making users revisit and recommend a website. For
complex decisions involving higher cognitive processes the impor-
tance of content design increases, whereas aesthetics have a lower
level of importance [
82
,
83
]. Stanford et al. studied the dierence
between consumers and experts. Consumers rely on website aes-
thetics when assessing trustworthiness of websites, unlike experts,
who rather focus on factors related to content design [79].
For content design it is important to note that it is application-
and domain-specic which kind of information is perceived as
useful and comprehensive. Approaches to generalize ndings about
content design across domains have resulted in ambiguous ndings,
which can be seen at the example of images as content type. On
the one hand, scholars focusing on social presence cues state that
trustworthiness can be increased by using images as they create a
feeling of face-to-face interactions [
25
,
70
,
81
]. On the other hand,
other studies could not nd an eect of images on trustworthiness
of user interfaces [69, 70].
Drawing from these ndings we consider content design to be
the most relevant factor for building trustworthy interfaces for
carbon oset projects. Given the domain-specic nature of content
design, this in turn highlights the importance of exploring what
constitutes useful and comprehensive information in the realm of
voluntary carbon osetting.
2.2.3 Measuring Trustworthiness of User Interfaces. There are sev-
eral approaches measuring trustworthiness in online contexts [
34
,
71
]. We build on a measurement instrument developed by McKnight
et al. that addresses trust as a multi-dimensional concept speci-
cally in the context of websites [
58
]. Their instrument is based on
four trust-related constructs, namely the institution-based trust,
the disposition to trust, trusting intentions, and trusting beliefs
[
58
]. For the context of this work, we focus on the construct of
trusting beliefs (competence, benevolence, and integrity). Various
HCI researchers used and adapted the measurement instrument
by McKnight et al. to measure perceived trustworthiness in online
contexts [
11
,
13
,
74
]. Focusing on website design, Secker et al. found
that high integrity and competence lead to higher trustworthiness,
and an untrustworthy experience is related to missing benevolence
and integrity [
74
]. Hence, benevolence and integrity are necessary
yet not sucient to establish trustworthiness on their own. For
websites to be perceived as trustworthy competence is need as well.
These ndings underline that perceived trustworthiness should be
measured by including all three constructs, as the combination of
benevolence, integrity, and competence has proven to be the most
reliable way to derive a holistic trust measurement.
It is important to note that the perceived trustworthiness of an
interface does not say anything about whether the carbon oset
project represented in the interface is actually deserving of trust. We
acknowledge that by exploring how to build trustworthy interfaces
in the context of carbon osets, there is a risk that the results can
be used to apply manipulative design techniques. Manipulative
design and dark patterns are, unfortunately, a reality and have been
a subject of interest at DIS (see e.g. [
7
]). However, we can only
detect, stop and prevent manipulation if we know what elements
can be used to manipulate. We believe that publishing open and
accessible research on trustworthy interfaces is thus ultimately of
public interest.
2.3 Charitable Giving
Charitable giving refers to any act of providing money, time, or
goods with no value or monetary return for the donor [
88
]. Donors
face a dilemma as they have to trust a charitable organization while
not being fully able to monitor their performance and quality [
67
].
The actions of charities and outcomes of donations are perceived as
highly intangible [
65
]. This phenomenon can also be observed with
carbon osets, as they also face a missing tangible counter value at
which a user can verify that a transaction succeeded. Interest within
the HCI community in the eld has picked up (e.g. [
4
,
22
,
86
,
87
])
as donations are increasingly performed and reported online and
donors increasingly demand more transparent communication.
To appear in DIS ’22, June 13-June 17, 2021, Virtual Event, USA Guzij et al.
2.3.1 Trust in Charitable Giving. Trust research in the eld of char-
itable giving focuses on contextualizing and recognizing trust in
the relationship between individuals and charitable organizations
[
36
,
92
]. Research found that traditional factors such as brand and
reputation, familiarity, and geographic distance strongly inuence
donations [
24
,
35
,
73
]. However, studies show that there is a shift
in requirements on what makes charities trustworthy. With media
reports on charity scandals claiming donations not being used for
their cause, the public increasingly demands greater accountability
and transparency [
24
]. Donors face a dilemma as they have to trust
a charitable organization while not being fully able to monitor their
performance and quality [
67
]. Donor questionnaires increasingly
state that charitable organizations should go beyond the (legal)
required minimum of reporting to increase public trust [
91
]. Yang
and Northcott derived that the public asks explicitly for direct ac-
countability through continuous communication to donors [
91
].
Sargeant and Lee also validated this measure [
73
]. Next to nancial
transparency, non-nancial information has been identied as crit-
ical to improve trustworthiness [
6
,
73
]. Non-nancial data can be
information on the outputs and outcomes that a charity achieved.
While outputs are dened as the direct result of the input, e.g. num-
ber of fed children, outcomes are dened as the indirect benets or
changes in, e.g. condition, status, or skills of the beneciaries [
16
].
2.3.2 Trustworthiness of Charitable Giving in Online Contexts. As
charitable giving is increasingly performed online, charities are
faced with the challenge of establishing trust cues in their online
communication. As online contexts allow anyone to publish any
information, traditional credibility cues such as the reputation are
not sucient to assess a charitable organization’s trustworthiness.
The increasing demand for transparency and online communica-
tion have thus started to motivate the HCI community to analyze
online representation of charitable projects. Yang and Northcott an-
alyzed how charities communicate their information and identied
missing comprehensiveness of communicated data [
91
]. Krasteva
and Yildirim examined the process of getting informed before and
after a donation [
49
]. They found that the main reason why only
a few people research before donating lies in the eort to retrieve
and process information. They highlight the importance of access
to data as there is an increasing number of donors for whom it is
not sucient to donate based on the feeling of "warm glow" – a
better emotional state of mind after donating. With these "rational
donors" in mind, the authors recommend charities to improve their
website communication [49].
In the context of website design, scholars studied trustworthiness
of charitable organizations based on the elements of visual design,
ease of use, and content design of their websites. Robins et al. found
that visual design judgments play a highly signicant role in rating
charities’ credibility [
72
]. Küchler et al. examined the eect of
website design perceptions on willingness to donate [
50
]. They
created an extension of the technology acceptance model by van
der Heijden and investigated the inuence on a user’s intention
to donate [
20
,
50
]. Their ndings state that content design has
the highest impact on willingness to donate. These results are in
line with ndings of Thielsch et al. arguing that content design
increases in importance when donating one’s own money compared
to someone else’s, as higher cognitive processes are being activated
[
83
]. These ndings explain why choosing the right type of content
becomes increasingly important for charitable organizations to
improve trustworthiness of their projects. Based on these ndings,
we assume an equally important role of content design to create
trustworthy interfaces for carbon oset projects.
2.4 Summary
Voluntary carbon oset projects have the potential to contribute to
the ght against climate change. As most communication on the
progress and impact of oset projects happens online, this poses
the challenge of designing user interfaces that facilitate trust. Draw-
ing on HCI literature we hypothesize that trust can be facilitated
through online user interfaces if they are perceived as trustworthy
(competence, benevolence, and integrity). Specic to online inter-
faces we know that visual design, content design and ease of use
inuence users’ perceived trustworthiness of websites. With many
parallels to voluntary carbon osetting, research about website
design of charitable organizations oers a starting point to explore
the design of trustworthy carbon oset interfaces. Both elds ad-
dress users that give voluntarily and the impact of transactions is
realized in distant locations or over long time periods. This makes
it dicult to verify whether trust was not misused. Research in
the eld of charitable giving identied content design – provid-
ing information that is perceived both useful and comprehensive
– as the most inuential factor for establishing trustworthiness
and inuencing the willingness to donate. Based on these ndings,
we assume an equally important role of content design to create
trustworthy interfaces for carbon oset projects. As content design
is highly domain specic, the main goal of this paper is to explore
which types of content can serve as reliable trust cues and how to
present them in user interfaces of carbon oset projects.
3 RESEARCH APPROACH
To determine potential trust cues for carbon oset interfaces, we
rst conducted semi-structured interviews to examine the buyer-
seller trust relationship and understand requirements of both sides.
We then sampled websites of existing carbon oset sellers to un-
derstand the existing design space. While the results of these steps
are not at the core of this paper, we think it is worth reporting how
they contributed to the establishment of our hypotheses.
3.1 Interview Study
We conducted semi-structured interviews with oset buyers (n=7)
and sellers (n=4) to examine the buyer-seller trust relationship and
understand requirements for communicating progress of carbon
oset projects, with specic focus on forest carbon osets. Our goal
was to build an initial qualitative understanding of the space, its
stakeholders, and their relationships. Grounded Theory provided a
systematic approach for simultaneous data collection and analysis
with a low risk of including researchers’ bias [
33
]. The small sample
size naturally limits the generalizability of these ndings. From the
interview data we identied three main challenges related to the
trust relationship between buyers and sellers, namely (1) choosing
a trusted seller for buyers, (2) providing transparent data on the
progress of carbon oset projects, and (3) communicating updates
in an easily understandable and accessible way. We identied four
Designing Trustworthy User Interfaces for the Voluntary Carbon Market To appear in DIS ’22, June 13-June 17, 2021, Virtual Event, USA
areas in which data can be used to communicate the progress of
oset project in a more transparent way:
Geographic focus
: Information on the exact location and size
of the carbon oset projects, visualization of the project area, e.g.
through satellite images.
Financial transparency:
Information on costs involved in car-
bon oset projects, including a split of costs and the dierent
stakeholders getting paid through the donation.
Forest-related measurements:
Continuous and easy-to-understand
updates on additionality (a comparison to a baseline scenario in
which the investment had not happened), accumulated biomass,
biodiversity measurements, and a clear measure of how much
carbon is stored in the project.
Images
: Images (or videos) of updates about the project. A par-
ticular interest was set on images of the forest area, tree planting
activities portraying workers, as well as evidence of the tree
growing progress.
3.2 Design Space Analysis
To better understand the existing design space for forest carbon
oset user interfaces, we additionally sampled websites of (n=7)
forest carbon oset sellers (see supplementary material). There are
a variety of design approaches for communicating project progress,
using dierent data types and levels of detail. User interfaces of
established organizations (e.g. ClimatePartner, South Pole) tend
to provide mainly qualitative data with little detail. Interfaces in-
cluded text-based descriptions, annual carbon oset capacity, the
area’s contribution to the Sustainable Development Goals (SDGs),
and images or videos providing context around the project. Images
typically portrayed dierent stages of forest growth, dierent types
of plants or habitats of animals, and sometimes plantation activities,
with a focus on portraying the inclusion of local communities. Only
few providers focused on more detailed approach, using quantita-
tive data to communicate project progress over time. Those who
did were startups (e.g. Ecologi, Pachama) rather than established
incumbents and they focused either on forest-related data or nan-
cial data. Interestingly, quantitative data was not always integrated
into the user interface of the specic project website. For example,
Ecologi redirected to a hosted spreadsheet to provide access to
the raw nancial data set. Figure 1 shows an overview of design
approaches.
3.3 Conceptual Framework
The results of our interview study and design space analysis indicate
a demand for and a trend towards interfaces that use quantitative
data to provide transparent updates on projects. While interfaces of
incumbents rely largely on qualitative data (e.g. images) to inform
about projects, emerging companies are trying to integrate quanti-
tative data (e.g. nancial data and forest data) to build trust through
transparent updates. This trend is interesting, because research in
online donations found that content design – i.e. what and how
to present content in interfaces – to be particularly important for
cognitive actions that go beyond the rst impression [
50
]. Building
these insights, we identied (1) the type of data and (2) the level
of detail at which to present it to be interesting dimensions to be
investigated as trust-building factors.
High detail
Low detail
QuantitativeQualitative
Photos
Sustainable
Development
Goals
Community
impact reports
Mentioning climate
measures taken in
project
Tree planting locations
Location data
Satellite data
Project timeline
Certification
(mention-only)Retired carbon
amount per month
Cumulative
amount of trees
planted
Relative split of
financial contribution
per donation
CO2 storage per
tree (species)
Biomass index Historical carbon
graphs (additional
carbon stored)
Financial reports
& receipts
Figure 1: The design space for online interfaces in the volun-
tary forest carbon market. While more recent companies ex-
periment with quantitative approaches, incumbents mainly
rely on qualitative data.
Type of Data. We theorize that the type of data portrayed on a
user interface impacts the selection of carbon oset projects and
their evaluation of trust. We operationalize this variable through
three instances: nancial data, images, and forest data. Financial
data transparency emerged throughout the interview study, design
space analysis, and literature from charitable giving [
6
,
73
,
91
]. Like-
wise, images were both mentioned by interviewees and are widely
used in existing interfaces. Prior HCI research, however, is incon-
clusive on their eect on trust, making them an interesting target
to investigate [
69
,
70
,
81
]. Finally, quantitative forest data emerged
in both the interview process and the design space analysis.
Level of Detail. Our analysis of existing interfaces shows that
there are few examples that visualize quantitative data at a high
level of detail. If data is provided at high level of detail it was not
well-integrated into the interface. The interview study indicates
that buyers wish for more detailed information, specically on
quantitative data. At the same time, sellers in our interview study,
aiming to provide more transparency, assumed limited interest in
quantitative data from buyers. These contradictory assumptions
raise the question of whether the level of detail aects project choice
and perceived trustworthiness. Thus, it is interesting to evaluate
how to integrate highly detailed quantitative data into the user
interface. As content design focuses on creating trust by providing
information that is perceived as useful and comprehensive, we
see the level of detail as a decisive dimension in designing user
interfaces for carbon osets.
3.4 Hypotheses
We theorize that the type of data (nancial, image, forest) and the
level of detail at which it is presented (low, high) in user inter-
faces inuence the trustworthiness users perceive towards a forest
carbon oset project. As dependent variables we investigate both
project choice by participants in a ctitious scenario as well as the
perceived trustworthiness. A detailed description of the dependent
variables can be found in the Method section.
To appear in DIS ’22, June 13-June 17, 2021, Virtual Event, USA Guzij et al.
Focusing on the type of data as independent variables we dene
the following hypotheses:
Hypothesis 1:
The type of data impacts the choice of a forest carbon
oset project.
Hypothesis 2:
The type of data impacts the evaluation of trust in a
forest carbon oset project.
Additionally, we suggest that the level of detail of the portrayed data
will impact project choice and perceived trustworthiness. Therefore,
we posit the following hypotheses:
Hypothesis 3:
The level of detail impacts the choice of a forest carbon
oset project.
Hypothesis 4:
The level of detail impacts the evaluation of trust in a
forest carbon oset project.
4 METHOD
We conducted a randomized online experiment with an interactive
prototype to test the eects of the (1) type of data (nancial, image,
forest) and (2) level of detail (low, high) on the project selection and
perceived trust. The following section describes our method.
4.1 Participants
We recruited 244 participants via university mailing lists, reaching
approximately 7000 recipients and sharing it on social media. The
entire experiment was conducted in English, and participants were
briefed about an approximate completion time of 10 minutes. To
increase participation, we committed to donating 0.20 EUR per
completed survey to an osetting project and raed three vouchers
worth 20 EUR among participants. Our nal sample consisted of 244
participants, out of which 54.1% were male, and 44.3% were female.
The average age of our participants was 27.41 years. 95.1% of our
participants lived in anonymized. Our participants were mainly
students (59.8%), followed by employees (32.4%), self-employed
people (6.1%), and retirees (1.6%).
4.2 Apparatus
4.2.1 Independent Variables. Based on the denition of our inde-
pendent variables and our hypotheses, we designed seven dierent
user interfaces using Figma
1
and deployed them as interactive web-
site with Useberry
2
. Each interface described a forest carbon oset
project with text-based project information (title, location, project
description). Six out of seven interfaces showed additional infor-
mation about the project by providing either nancial, image, or
forest data at low or high levels of detail. We were careful not to
include any other dierence in the interface or project description
that could inuence the experiment. Figure 2 shows screenshots of
the six interfaces. The control project and overview page are not
depicted, but can be found in the supplementary material.
1https://www.gma.com/ (last accessed 2021-08-09)
2https://www.useberry.com/ (last accessed 2021-08-09)
Interface 1A/ 1B: Financial Data. Financial data provides addi-
tional information on how each Euro invested in the project is
spent. We communicated that 40% of the donations are needed for
administrative overhead, and 60% are invested into reforestation
activities and visualized the information as an annotated donut
chart. In the low level of detail interface (1A), this was the only in-
formation provided. The high level of detail interface (1B) portrays
a more granular split with direct nancial indication to the dier-
ent cost blocks of the 40/60 split. The relative cost block indication
was derived from research on cost splits of the clean development
mechanism [68].
Interface 2A/2B: Images. We analyzed images of existing providers
to identify commonly used image content. We portray the progress
of tree planting by showing images of plants in dierent types of
progress and included plantation workers in the images. We did
not include images that might trigger strong emotional responses,
such as direct-facing portraits, families, or children. In the low level
of detail interface (2A), one image is shown. The high level of detail
interface (2B) shows additional images to provide more context.
Interface 3A/3B: Forest Data. Forest data refers to quantitative
information on the development of the specic forest within the
project. We focused specically on the number of trees planted and
the annual tons of carbon oset. In the low level of detail interface
(3A), the total of planted trees and hectares of forest protected
and the annual carbon oset are shown. The high level of detail
interface (3B) shows additional information about the type of trees
planted, a biodiversity index, and a comparison graph modeling
the annual amount of oset carbon with and without the project
from its beginning to 2060.
Interface 4: Control. To create a standard comparison basis for
the experiment, we took inspiration from the descriptions of the
sampled carbon oset sellers and created a control project. All ele-
ments of the control project are also present in the other interfaces.
We altered only the structure and formulation of sentences so that
participants did not feel reading the same text across all projects.
We were careful to describe the project in a similar language and
format as found for real projects. The interface describes the project
with a title, project id, location name, and a tag on a map. A text-
based description explains the project setup, providing a qualitative
report summary as typically done at established carbon oset sell-
ers. It consists of a problem statement and a description of the
project’s activities, with specic mentions that the project benets
biodiversity or local community education. To avoid bias through
geographical proximity we located all projects in the Amazon Rain-
forest region of South America, expecting our participant sample
to come predominantly from anonymized [35].
4.2.2 Dependent Variables. The two dependent variables were the
selected project and perceived trust in each project.
Project Selection: After reviewing all interfaces, each participant
had to select one project in which to invest. We measured the project
selection as a dichotomous variable for each project [0=No;1=Yes].
Perceived Trust: After selecting one project, participants were
asked to rate each project on a ve-point Likert scale. Following the
example of previous research on trust in online contexts [
11
,
13
,
74
]
we adapted three questions from the Trusting Beliefs instrument
Designing Trustworthy User Interfaces for the Voluntary Carbon Market To appear in DIS ’22, June 13-June 17, 2021, Virtual Event, USA
Figure 2: The prototyped interfaces. Each interface shows a general text-based project description on the left side and addi-
tional data on the right side. The additional data operationalizes the independent variables, namely the type of data and the
level of detail at which it is presented. The rst column from the left shows nancial data, the second one images, and the third
one forest data. The rst line shows the low level of detail and the second line shows the high level of detail. High resolution
images are in the supplementary material.
by McKnight et al. [
58
]. To acknowledge the multi-dimensionality
of trust, the averaged scores of the three questions represented the
nal trust value index.
(1) I believe this project would act in my best interest.
(2) I would characterize this project as honest and trustworthy.
(3)
This project performs its role of communicating updates on their
progress very well.
4.3 Experimental Design
The design of our experiment uses a split-plot setup, utilizing both
between-group and within-group components [
52
]. To investigate
the inuence of the data type, we adopted a within-group design –
each participants was shown interfaces with with each type of data
(nancial, image, forest data, control). To investigate the inuence
of the level of detail, we adopted between-group design – each par-
ticipant was shown all interfaces either from the low-detail (Group
A) or from the high-detail (Group B) condition. Table 1 shows the
dierent experimental conditions. As both groups also contain a
control project, the total number of conditions in this experiment
is seven. Each participant was shown exactly four interfaces (one
for each data type). To avoid sequence and carryover eect for the
within-group design, we adopted a Latin square. The order of the
conditions was systematically arranged in a counterbalanced order,
and participants were randomly assigned to the conditions [
52
]. We
were careful to avoid any dierence between the interfaces aside
from the independent variables investigated (see supplementary
material for screenshots of all interfaces).
Table 1: Overview of the experimental conditions. Partic-
ipants saw the control project and either interfaces from
Group A or B.
within group
Financial Images Forest Control
between
group
low detail 1A 2A 3A 4
high detail 1B 2B 3B
4.4 Procedure
The experiment was administered online with the survey tool Use-
berry
3
, allowing interactive website-like interaction with the proto-
typed interfaces. The data was collected anonymously. The experi-
ment only enabled participation for desktop devices. Participants
in total had to complete six steps: (1) answer a brief entry survey,
(2) read the scenario description, (3) browse through four projects,
(4) select one project, (5) evaluate the trust of each project, and (6)
answer a demographic survey. The scenario provided to partici-
pants stated that they should imagine being customers of a bank
committed to becoming climate-neutral. As customers, they are
asked to select in which project the bank should invest on their
behalf. Given recent media reports, we asked participants to pay
attention about choosing a trustworthy project.
5 RESULTS
In total 244 participants completed the experiment. Given the split-
plot design, our sample can be divided into two groups: Group A,
which saw the low detail versions of the projects (n=119), and Group
B, which saw the high detail versions of the projects (n=125). The
3https://www.useberry.com/ (last accessed 2021-08-09)
To appear in DIS ’22, June 13-June 17, 2021, Virtual Event, USA Guzij et al.
sample in both groups mainly consists of students (58% & 61.6%)
with a similar average age (28, SD=8.228 & 27.5, SD=7.139). The
average completion time of the experiment was 9.25 minutes.
5.1 Behavior of Overall Sample
Participants’ project choices show dierences between the low
detail and high detail sample. Table 2 shows an overview of the
selected projects by group. Both groups decided on Project 3 (for-
est) with the highest proportion (55.5%, 43.6%). Project 1 (nancial)
scored the second. While participants in the low detail group se-
lected Project 1 24.4% of the time, the proportion almost doubles to
40% in the high detail group. Project 2 (image) ranks third in the
high detail group (9.6%). In the low detail group, however, it was
selected only 8.2% of the time – less often than Project 4 (control).
Project 4 (control) was selected 11.9% in the low detail group and
4.8% of the time in the high detail group.
Table 2: Overview of the project selection. The relative fre-
quencies at which projects were selected by Group A (low
detail) and Group B (high detail)
Group A Group B
(low detail) (high detail)
Project 1 (Financial) 24.4% 40.0%
Project 2 (Images) 8.2% 9.6%
Project 3 (Forest) 55.5% 43.6%
Project 4 (Control) 11.9% 4.8%
After selecting a project, we asked participants to evaluate the
trustworthiness of each project. Figure 3 provides an overview of
participants’ trust index ratings. The results show that Project 3 (for-
est), which was also selected most often, scored the highest across
both groups (Mode 3.3 & 3.7; Mean 3.7 & 3.4). Project 1 (nancial)
is perceived more trustworthy in the high detail group. For Project
2 (image) there is a strong deviation between mode and mean in
Group A (2.7 & 1.66), resulting in a left-skewed distribution that
indicates discrepancy in opinions regarding the trustworthiness.
Project 2 (image) showed no dierences between groups, while
Project 4 (control) is evaluated with lower trust in Group B. Here,
we also see a left-skewed distribution (2.3 & 1.66).
5.2 Hypothesis Testing
The collected data consists of dierent types of variables. Since all
of them are ordinal measures, they do not meet the assumptions
of parametric tests. Therefore, we tested our hypotheses with non-
parametric tests [52].
5.2.1 Type of data & choice of project: H1 deals with the question
of whether the type of data [nancial, image, forest] impacts the
choice of forest carbon oset projects. We conducted the Cochran’s
Q test [14]. It compares more than two dependent variables based
on the ranks of the dependent variable for signicant dierences
to verify whether ktreatments have identical eects. The data
comparison was made within groups 1 and 2 separately.
Group A (low detail). There is a statistically signicant dierence
between the types of data,
-2
(3)=65.639, p<.001. As the Cochran
Q test only states the existence of a dierence, a post hoc test
determined the exact dierence between the projects. There is a
statistically signicant dierence between Project 2A [images] and
0% 100%100% 50%75% 25% 50%75% 25%
Group A(low detail)
Project 1 (Financial)
Project 2 (Images)
Project 3 (Forest)
Project 4(Control)
1 Strongly Disagree 2–Disagree 3 – Neutral 4–Agree 5–Strongly Agree
0% 100%100% 50%75% 25% 50%75% 25%
Group B(high detail)
Project 1 (Financial)
Project 2 (Images)
Project 3 (Forest)
Project 4(Control)
1 Strongly Disagree 2–Disagree 3 – Neutral 4–Agree 5–Strongly Agree
Figure 3: The trust index for each project by group. Partic-
ipants rated the trustworthiness of each project by answer-
ing three questions on a ve-point Likert scale. A trust in-
dex was calculated as the mean value of the three answers.
The diverging stack bar charts provide an overview of partic-
ipants’ responses by group and project. Comma values were
rounded to generate the graph.
3A [forest] (p<.001); Project 4 [control] and 3A [forest] (p<.001);
as well as Project 1A [nancial] and 3A [forest] (p<.001). Project
2A [images] and 4 [control] (p<.604); Project 2A [images] and 1A
[nancial] (p<.014); likewise Project 4 [control] and 1A [nancial]
(p<.052) showed no statistically signicant dierence. Our pairwise
comparison shows that Project 3A [forest] was the leading indicator
for the dierence in Group A.
Group B (high detail). There is a statistically signicant dier-
ence between types of data,
-2
(3)=64.729, p<.001. The post hoc
test shows dierent results in the pairwise comparison of group 2.
There is a statistically signicant dierence between Project 4 [con-
trol] and 1B [nancial] (p<.001); Project 4 [control] and 3B [forest]
(p<.001); Project 2B [images] and 1B [nancial] (p<.001); as well
as Project 2B [images] and 3B [forest] (p<.001). Project 4 [control]
and 2B [images] (p<.448); likewise Project 1B [nancial] and 3B
[forest] (p<.376) showed no statistically signicant dierence. In
summary, we can state the following:
Group A: Reject 10, Retain 11
The type of data [nancial, image, forest] impacts the choice of
a forest carbon oset project.
Group B: Reject 10, Retain 11
The type of data [nancial, image, forest] impacts the choice of
a forest carbon oset project.
5.2.2 Type of data & evaluation of trust: H2 deals with the question
of whether the type of data [nancial, image, forest] impacts the
Designing Trustworthy User Interfaces for the Voluntary Carbon Market To appear in DIS ’22, June 13-June 17, 2021, Virtual Event, USA
evaluation of trust in forest carbon oset projects. This hypothe-
sis was initially tested within-group, using Friedman’s two-way
analysis of variance, a non-parametric alternative to the one-way
repeated measures ANOVA [
52
]. It checks for a dierence in multi-
ple dependent samples and the population from which these have
been drawn [
52
]. If the Friedman test is statistically signicant,
we can say that there is at least one dierence between one of
the variables. Our results show a statistically signicant dierence
between the control project, nancial data, forest data, and im-
ages when evaluating the trust of a carbon oset project,
j2
(3) =
325.471, p <.0001. The post hoc analysis shows that the pairwise
comparison is statistically signicant between all projects, except
between Project 2 [images] and Project 1 [nancial], which are not
statistically signicant (p <0.12).
As these results dier from the project choice analysis, we con-
ducted a second round of Friedman tests separating Group A and
Group B. The separate analysis shows that both groups would re-
ject the null hypothesis (p < .001). However, the post hoc pairwise
comparisons show dierences between the groups. Group B, the
high detail group, is dierent across all pairwise comparisons with
statistical signicance. Hence, in the high detail group the type of
data makes a dierence when evaluating the trust of a forest car-
bon oset project. Group A, the low detail group, shows the same
results as the cross-group analysis. All pairwise comparisons show
statistically signicant dierences, except for Project 2 [images]
and Project 1 [nancial], which achieve no statistical signicance in
their pairwise comparison (p < .422). The result of this in between-
group dierences will be addressed in H4. In summary, we can state
the following:
Reject 20, Retain 21
The type of data impacts the evaluation of trust in a forest carbon
oset project.
5.2.3 Level of detail & choice of project. H3 deals with the question
of whether there is a dierence in low level and high level of detail
when users choose their preferred oset project. We performed
a between-group analysis to analyze frequency counts between
Group A and Group B. We used the chi-square test of homogene-
ity
j2
[
80
]. There was a statistically signicant dierence in the
multinomial probability distributions between the two groups (p
= .024). Therefore, we can reject the null hypothesis and accept
the alternative hypothesis. Our cross-tabulation results show that
people in Group B choose Project 1 [nancial] (13.7% vs. 63.3%)
more often than in Group A. People in Group B also choose Project
3 [forest] less often than in Group A (53.7% vs. 46.3%). Project 2
[images] was selected more often in Group B (45.5% vs. 54.5%).
Project 4, our control group, was selected signicantly less often in
Group B (70% vs. 30%). In summary, we can state the following:
Reject 30, Retain 31
The level of detail impacts the choice of a forest carbon oset
project.
5.2.4 Level of detail & evaluation of trust: H4 assumes in its
0
that
the level of detail does not impact the evaluation of trust in a forest
carbon oset project. Our dependent variable is the trust index,
calculated as mean of the three trust questions answered on an
ordinal ve point Likert scale. Each project is analyzed separately.
The level of detail (1 for low, 2 for high) represents our dichotomous
independent variable. Statistical analysis does not directly test for
association between a dichotomous independent variable and an
ordinal dependent variable [
8
,
80
]. However, research suggests run-
ning an ordinal logistic regression and using the coecient (of the
dichotomous variable) from this analysis to establish a statistically
signicant relationship between these two variables using an odds
ratio [
52
,
80
]. The odds of an event occurring is the probability of
it occurring versus the probability of it not occurring.
Project 1 (Financial). Before looking at the result of the odds
ratio for project 1, it is necessary to check the overall model t, i.e.,
whether the data sample is appropriate for the regression model
[
80
]. The Pearson as well as the deviance goodness-of-t test indi-
cated that the model was a good t to the observed data,
j2
(11) =
10.573, p < .480,
j2
(11) = 11.537, p < .399. The likelihood-ratio test,
also called the model-tting information, evaluates the change in
model t by comparing the full model to the intercept-only model.
The nal model statistically signicantly predicted the dependent
variable over and above the intercept-only model,
j2
(1) = 10.634,
p < .001. This means that the independent variable adds to the
prediction of the dependent variable. The odds of seeing low level
of detail and rate a high trust value was 0.476 (95% CI, 0.304 to
0.745), a statistically signicant eect,
j2
(1) = 10.537, p < .001. Since
the regression weight (B score) shows a negative value (-0.742),
this means that people who have a low detailed prototype are not
likely to evaluate Project 1 with a high trust score. The higher
the regression weight, the likelihood of rating a high trust value
increases.
Project 2 (Images). The Pearson as well as the deviance goodness-
of-t test indicated that the model was a good t to the observed
data,
j2
(11) = 12.185, p < .350,
j2
(11) = 13.323, p < .273. The
likelihood-ratio test did not achieve the desired statistical signi-
cance,
j2
(1) = 162, p < .688. This result means that the independent
variable does not account for the prediction of the dependent vari-
able. The odds of seeing low level of detail and rate a high trust
value was 0.914 (95% CI, 0.590 to 1.416), a not statistically signi-
cant eect,
j2
(1) = 0.162, p < .687. Although the regression weight
shows a negative value (-.0.90), we do not have signicant results to
state that a low detailed prototype can predict a high trust score for
Project 2. Thus, it does not make a dierence if it is a low detailed
or high detailed prototype for the trust evaluation of Project 2.
Project 3 (Forest). The Pearson as well as the deviance goodness-
of-t test indicated that the model was a good t to the observed
data,
j2
(9) = 14.086, p < .119,
j2
(9) = 15.059, p < .089. The likelihood-
ratio test did not achieve the desired statistical signicance
j2
2(1)
= 0.162, p < .813. This result means that the independent variable
does not account for the prediction of the dependent variable. The
odds of seeing low level of detail and rate a high trust value was
1.055 (95% CI, 0.678 to 1,640), a not statistically signicant eect,
j2
(1) = 0.56, p < .813. Even if the regression weight shows a positive
value (.053), we do not have signicant results to state that a low
detailed prototype can predict a high trust score for Project 3. Thus,
it does not make a dierence for the trust evaluation of Project 3 if
it is a low detailed or high detailed prototype.
To appear in DIS ’22, June 13-June 17, 2021, Virtual Event, USA Guzij et al.
As two projects showed a dierence in level of detail when evaluat-
ing trust, we can summarize the following:
Reject 40, Retain 41
The level of detail impacts the evaluation of trust in a forest
carbon oset project.
6 DISCUSSION
This study set out to explore how to design trustworthy interfaces in
the context of the voluntary carbon oset market. Our experimental
results show that the choice and the evaluation of trustworthiness in
a forest carbon oset project depends on the type of data portrayed
in the user interface and its level of detail. We found that not all data
types are equally eective as trust cues. Additionally, the right level
of detail must be assessed for each data type separately, as a higher
level of detail does not automatically increase trustworthiness for
each data type.
Our study results are in line with research from the eld of
charitable giving, where content design, with specic focus on
quantitative data shows to have the strongest inuence on perceived
trustworthiness of charitable projects. This study is the rst to
report specic guidance for interface designers on what data may
serve as trust cues in the domain of carbon osets. More generally,
our ndings shed new light on the question of how designers can
create online experiences that facilitate trust through interfaces in
situations where users cannot check the fulllment of a transaction
directly. In the following we discuss both theoretical an practical
implications for interaction design.
6.1 Overview
We identied quantitative forest data as the strongest predictor
for evaluating a forest carbon oset project as trustworthy and
ultimately choosing it. Our pairwise comparisons with all other
data types showed that participants selected projects reporting
forest data signicantly more often. A low level of detail, namely
showing the protected area and the annual oset volume was al-
ready sucient to establish trust into the project. Projects that
provided nancial transparency were chosen the second most of-
ten and ranked second in the evaluation of trustworthiness. The
level of detail at which data was made available made a signicant
dierence for interfaces showing nancial data. By providing more
detail, the proportion of people selecting the project almost dou-
bled from 24.4% in Group A to 40% in Group B. This result, for one,
highlights the positive impact of nancial transparency on trust-
worthiness and conrms ndings from the eld of charitable giving
[
6
,
73
,
91
]. However, it also sheds light on the fact that establishing
trust through user interfaces depends on not only the type of data,
but also how it is presented. Projects providing images portraying
the forest area or tree planting activities did not aect trustworthi-
ness or project choice. This is in line with literature, nding that
images can create a broad range of reactions, from suspicion to
enthusiasm [
69
]. Our data indicates that the images we used did
not aect trust in the context of forest carbon osets. However,
images with dierent content – i.e., emotional content, progress
of a specic tree, pictures contextualized with specic captions –
might lead to dierent results. Given contradictory ndings from
literature, it could be interesting to revisit the impact of images and
other qualitative trust cues in combination with quantitative data,
as they could improve trustworthiness as a complementary data
source.
The most relevant question in the context of this paper probably
is the following: Does more transparency and providing more data
in user interfaces lead to increased trust in forest carbon oset
projects? In summary, our results show that we indeed can facili-
tate trust through online interfaces. Our results also point to the
importance of contextualizing content design. It is crucial to under-
stand which data is essential within the relevant domain context
and how to present it. In the following, we discuss and present
both theoretical and practical implications for HCI and interaction
design that arise from these ndings.
6.2 Theoretical Implications
From our results we can draw several theoretical implications.
The increasing demand for voluntary carbon credits and lack of
transparency in today’s market will lead to increased pressure to pro-
vide more transparent and continuous data on carbon oset projects.
The voluntary carbon oset market is facing accelerated growth in
demand. While established companies rely on their brand reputa-
tion, new market entrants will need to nd ways to compete with
them for consumers’ trust. From our interview study and design
space analysis, we can speculate that new oset companies will use
more transparent and continuous data updates to establish trust
and by doing so create the brand value to challenge incumbents. At
the same time – with a similar development already today in chari-
table giving – we expect consumers to push for more transparency
of carbon oset projects. Given the missing standardization, we
suggest that the HCI community should pursue an active role in
further understanding and establishing guidelines on presenting
and integrating data into user interfaces. Along with technical in-
novations enabling better tracking of projects’ impact, it is equally
important to develop design standards to present the data so that
it can be easily understood and interpreted by users. Particularly,
the eld of Sustainable HCI has the potential to go beyond interac-
tions between humans and computers and include interdisciplinary
research to dene design approaches based on psychological and
economic decision models [3, 15, 46].
Integrating additional data into user interfaces of carbon oset
projects facilitates the establishment of trust in online contexts. Our
results show that providing additional data on carbon oset projects
supports the trustworthiness of these projects. Our results con-
nect to literature on charitable giving that highlights the impor-
tance of data transparency [
91
]. However, not all information is
equally suited to facilitate trust. For interfaces of forest carbon
oset projects, quantitative data on a project’s development, such
as forest or nancial data, outperforms qualitative data, such as
images. This advantage of quantitative data is likely to transfer to
other domains and projects, such as other types of carbon osetting,
plastic osetting, as it already created a shift in the eld of chari-
table giving in general. Given the domain-specic nature of trust
cues, it is necessary to identify what type of data is appropriate for
each context.
How data is presented in online interfaces impacts trustworthiness.
Our results show that it is not enough to only present the right
Designing Trustworthy User Interfaces for the Voluntary Carbon Market To appear in DIS ’22, June 13-June 17, 2021, Virtual Event, USA
type of data. It is also important to do it in the right way. Dierent
types of data are aected dierently by this relationship. For nan-
cial data, communication at a too supercial level does not result
in increased trustworthiness. Instead, it might rather raise users’
suspicions if the provided data appears not to be comprehensive.
For forest data, on the other hand, interface elements providing
low-detail information already had a positive eect. In such a case,
presenting data in a too detailed format may well decrease the
desired eect as it overloads users with data they are not familiar
with interpreting. Consequently, it is necessary to analyze each type
of data independently to identify the appropriate level at which
to present it. While our research focused on the level of detail to
investigate, it is likely that there are more dimensions (i.e. visual
design, ease of use) that are relevant.
6.3 Practical Implications
From our ndings we derive three implications for practitioners
and interface designers.
Experiment with continuous quantitative data in user interfaces:
Our study results show that quantitative data supports the establish-
ment of trust better than qualitative data. Literature from charitable
giving further highlights the importance of continuous updates to
increase donor lifetime value [
21
]. From this we can infer that users
will develop more trust, if they receive updates about the impact of
their carbon osets over time. Quantitative data is more dicult to
interpret for users, thus practitioners will need to meet this design
challenge through experimentation with dierent data types, their
combinations, and visualization. The results of our study suggest
that focusing on inputs, outputs, outcomes [
26
] related to nancial
and forest data may be practical starting points to communicate
continuous progress. Providing data in machine-readable formats
would allow integration into other products and services, which,
in turn, would help create transparency. We recognize that devel-
oping interfaces with frequently updated information comes with
many implementation challenges but want to highlight it as an
opportunity to build long-term brand value in a growing market.
Build interfaces that guide users’ sense-making process to establish
trust: Our research points to the importance of domain experts, i.e.
oset sellers, to closely collaborate with user interface designers to
nd approaches to visualize data appropriately. We highlight the
role of user interface designers in guiding users’ sense-making pro-
cess when being confronted with quantitative data on carbon oset
projects. Given the complexity of carbon osets, it is important to
nd the appropriate abstraction level to present dierent data types.
We suggest starting at a high abstraction layer that communicates
the most important metrics and allows users to understand the big
picture, but then allowing them to drill down to see more detailed
data if interested. At the example of nancial data interfaces this
can be realized in charts portraying high-level nancial information
of a project, allowing users to deep dive into specicnancial indi-
cators by interacting with the chart. Ultimately, such an approach
would support serving dierent user groups and allow to work
around trade-os such as between intuitive understandability and
comprehensiveness [45]. Finally, these ndings also underline the
importance of creating interactive interfaces that allow users to
access their desired data points from one single source, without
overloading them.
Design for dierent user groups: warm-glow and rational osetters:
From literature in charitable giving, we know that donors can be
distinguished into "warm-glow" donors and rational donors [
47
].
Similarly, people are likely motivated by dierent reasons to o-
set. From our pre-study and the ambiguous results of image data
interfaces, we speculate that there are both "warm-glow" osetters
and rational osetters. Whereas rational osetters are interested in
understanding the impact of their investment, warm-glow oset-
ters are motivated by the "feeling of a warm-glow" after investing.
It might be worth considering these target groups and their needs
independently when designing user interfaces. Drawing from re-
search in the eld of persuasive technologies for sustainable behav-
ior change, we believe that targeted design has additional potential
to create change behaviors and attitudes of users towards a more
rational and transparency-focused assessment of data [59, 60].
6.4 Future Research Directions
Like any empirical study, ours is not without limitations. Our sample
is not representative of the general anonymized population. More
importantly, we simulated the selection of the carbon oset project,
and, therefore, no real investment by participants had to be made.
In the following, we outline how future work may overcome these
limitations and expand on our work.
Field Study:
We propose to run a study in collaboration with
an established carbon oset seller, aiming for a representative
sample. For one, this would allow us to investigate consumers’
motivation to oset and evaluate our hypothesis in a real-world
setting. It would be interesting to further segment participants
to understand whether dierent groups of osetters exist. For
example, it would be interesting to understand whether prior
knowledge about oset markets impacts the perceived trustwor-
thiness of interfaces.
Further Trust Cues:
Our experiment is limited to three data
types and two levels of details. Future work might expand on
both dimensions to better understand trust cues for carbon o-
set projects. For example, certication labels and geographical
location might be good starting points.
Interaction Eects:
We investigated the eect of each data
type in isolation. Future work could address the interaction be-
tween dierent data types on perceived trust and project selec-
tion. Specically, the role of image data as a complementary data
source could be an interesting starting point.
Dynamic Data Exploration:
Drawing from the state of current
interfaces for carbon oset projects, we conducted the experi-
ment with static data presented in the interfaces. Interaction
design research could investigate the impact of interactive inter-
face elements that allow users to explore data themselves.
Veriable Data:
While this work looks into what data types and
what presentation facilitate trust, we do not address how users
might verify the correctness of the presented data [
45
]. The col-
laboration between many stakeholders makes this a sizable chal-
lenge. Future work might address this from both a user-centric
and a technical perspective. New technologies such as blockchain
might be well suited to provide transparency on a technical level
To appear in DIS ’22, June 13-June 17, 2021, Virtual Event, USA Guzij et al.
[
39
,
48
,
63
]. However, blockchain systems themselves remain
dicult to use for both beginners [27, 28] and established users
[
29
]. HCI research should explore how to build interfaces that
remove these barriers and allow users to verify their osets.
Manipulative Design and Dark Patterns:
With initial insights
on how content can be adapted to create trustworthy interfaces
for carbon osets, we acknowledge the associated risk of play-
ers with false motivation being enabled to create manipulative
designs and promote carbon osets that do not deliver upon
their promise. The application of HCI methods can be used to
understand the factors of content design that inuence users to
become victims of dark patterns and establish metrics on how to
work against them.
7 CONCLUSION
This paper examines which data users need to evaluate the trust-
worthiness of forest carbon projects, and investigates how this
data can be best translated into online user interfaces. Our work
was motivated by the growing voluntary carbon market and con-
sumer demand for more transparent communication. The results
of our randomized experiment reveal that by providing the right
data in the right way user interfaces can facilitate the establish-
ment of trustworthiness. They also show that not all data types are
equally eective as trust cues and the level of detail at which data
is presented in the interface is relevant depending on the data type.
In our experiment, integrating quantitative forest-related and de-
tailed nancial data in the user interface increased trustworthiness,
whereas images on their own had no statistically signicant eect.
Rooted in these ndings, we present both theoretical and practical
implications for researchers and designers. Doing so, this study is
the rst to report specic guidance for interface designers on what
data may serve as trust cues in the domain of carbon osets. On a
more general level, our research points to the importance of user
interfaces that guide users through complex data with appropriate
design, especially in cases where there is a high uncertainty if a
certain outcome has been achieved or not. We derive from our
ndings that contextual research in content design will become
increasingly important for HCI research in the eld of non-tangible
services. With this work we also hope to increase eorts enabling
oset buyers to better deal with information delivery on carbon
osets and identify manipulative designs. Whereas knowing about
domain-specic content design artifacts creates a starting point to
include additional mechanisms to create trust through transparency
in the eld of carbon osets.
ACKNOWLEDGMENTS
This work was supported by the Deutsche Forschungsgemeinschaft
(DFG) (grant no. 316457582 and 425869382). We thank the team
from https://condens.io/ for supporting us with their qualitative
research analysis tool — it helped us analyze and understand the
interview data we collected.
REFERENCES
[1]
Brian P Bailey, Laura J Gurak, and Joseph A Konstan. 2001. An examination
of trust production in computer-mediated exchange. In Proceedings of the 7th
Conference on Human Factors and the Web. 1–7.
[2]
Brian P Bailey, Laura J Gurak, and Joseph A Konstan. 2003. Trust in cyberspace.
Human factors and Web development (2003), 311–21.
[3]
Roy Bendor, Lisa P.Nathan, Matthew Louis Mauriello, and Oliver Bates. 2021. ’Ev-
erything in the Forest is the Forest’: A Decade of the Sustainability in (Inter)Action
Forum. Interactions 28, 4 (June 2021), 65–67. https://doi.org/10.1145/3466152
[4]
Nicola J. Bidwell, Chris Elsden, Ludwig Trotter, Josh Hallwright, Sadie Moore,
Kate Jeite-Delbridge, Mike Harding, Peter Shaw, Nigel Davies, Chris Speed, and
John Vines. 2021. A Right Time to Give: Beyond Saving Time in Automated
Conditional Donations. Association for Computing Machinery, New York, NY,
USA. https://doi.org/10.1145/3411764.3445371
[5]
Christopher Blaufelder, Cindy Levy, Peter Mannion, and Dickon Pinner.
2021. A blueprint for scaling voluntary carbon markets to meet the climate
challenge. https://www.mckinsey.com/business-functions/sustainability/our-
insights/a-blueprint- for-scaling-voluntary-carbon-markets- to-meet- the-
climate-challenge
[6]
Marie C. Blouin, Roderick L. Lee, and G. Scott Erickson. 2018. The impact of
online nancial disclosure and donations in nonprots. Journal of Nonprot
& Public Sector Marketing 30, 3 (July 2018), 251–266. https://doi.org/10.1080/
10495142.2018.1452819
[7]
Kerstin Bongard-Blanchy,Arianna Rossi, Salvador Rivas, Sophie Doublet, Vincent
Koenig, and Gabriele Lenzini. 2021. ”I Am Denitely Manipulated, Even When
I Am Aware of It. It’s Ridiculous!” - Dark Patterns from the End-User Perspective.
Association for Computing Machinery, New York, NY, USA, 763–776. https:
//doi.org/10.1145/3461778.3462086
[8]
Jürgen Bortz. 1999. Statistik. Springer Berlin Heidelberg, Berlin, Heidelberg.
https://doi.org/10.1007/978-3- 662-10031-8
[9] Josh Boyd. 2003. The Rhetorical Construction of Trust Online. (2003).
[10]
Mark Carney. 2021. TASKFORCE ON SCALING VOLUNTARY CARBON MAR-
KETS. https://www.iif.com/tsvcm
[11]
Luis V. Casaló and Jesús Cisneros. 2008. An Empirical Test of the Multiplicative
Eect of Usability on Consumer Trust and Satisfaction. In 2008 19th International
Conference on Database and Expert Systems Applications. IEEE, Turin, Italy, 439–
443. https://doi.org/10.1109/DEXA.2008.139
[12]
Sandy C. Chen and Gurpreet S. Dhillon. 2003. Interpreting Dimensions of Con-
sumer Trust in E-Commerce. Information Technology and Management 4, 2/3
(2003), 303–318. https://doi.org/10.1023/A:1022962631249
[13]
Jin-Hee Cho, Kevin Chan, and Sibel Adali. 2015. A Survey on Trust Modeling.
Comput. Surveys 48, 2 (Nov. 2015), 1–40. https://doi.org/10.1145/2815595
[14]
W. G. Cochran. 1950. THE COMPARISON OF PERCENTAGES IN MATCHED
SAMPLES. Biometrika 37, 3-4 (Dec. 1950), 256–266. https://doi.org/10.1093/
biomet/37.3-4.256
[15]
Rob Comber. 2017. Towards a social practice theory perspective on sustainable
HCI research and design. In Digital Technology and Sustainability. Routledge,
31–43.
[16]
Ciaran Connolly and Noel Hyndman. 2004. Performance reporting: a comparative
study of British and Irish charities. The British Accounting Review 36, 2 (June
2004), 127–154. https://doi.org/10.1016/j.bar.2003.10.004
[17]
Eric Corbett and Christopher Le Dantec. 2019. Towards a Design Framework
for Trust in Digital Civics (DIS ’19). Association for Computing Machinery, New
York, NY, USA, 1145–1156. https://doi.org/10.1145/3322276.3322296
[18]
Eric Corbett and Christopher A. Le Dantec. 2018. Exploring Trust in Digital
Civics. In Proceedings of the 2018 Designing Interactive Systems Conference (Hong
Kong, China) (DIS ’18). Association for Computing Machinery, New York, NY,
USA, 9–20. https://doi.org/10.1145/3196709.3196715
[19]
Fred D. Davis. 1989. Perceived Usefulness, Perceived Ease of Use, and User
Acceptance of Information Technology. MIS Quarterly 13, 3 (Sept. 1989), 319.
https://doi.org/10.2307/249008
[20]
Fred D. Davis, Richard P. Bagozzi, and Paul R. Warshaw. 1989. User Acceptance of
Computer Technology: A Comparison of Two Theoretical Models. Management
Science 35, 8 (1989), 982–1003. http://w ww.jstor.org/stable/2632151
[21]
A. S. Dick and K. Basu. 1994. Customer Loyalty: Toward an Integrated Conceptual
Framework. Journal of the Academy of Marketing Science 22, 2 (March 1994),
99–113. https://doi.org/10.1177/0092070394222001
[22]
Chris Elsden, Ludwig Trotter, Mike Harding, Nigel Davies, Chris Speed, and
John Vines. 2019. Programmable Donations: Exploring Escrow-Based Conditional
Giving. Association for Computing Machinery, New York, NY, USA, 1–13. https:
//doi.org/10.1145/3290605.3300609
[23]
Kate Ervine. 2012. The Politics and Practice of Carbon Osetting: Silencing
Dissent. New Political Science 34, 1 (March 2012), 1–20. https://doi.org/10.1080/
07393148.2012.646017
[24]
Megan M. Farwell, Micheal L. Shier, and Femida Handy. 2019. Explaining Trust
in Canadian Charities: The Inuence of Public Perceptions of Accountability,
Transparency, Familiarity and Institutional Trust. VOLUNTAS: International
Journal of Voluntary and Nonprot Organizations 30, 4 (Aug. 2019), 768–782.
https://doi.org/10.1007/s11266-018- 00046-8
[25]
BJ Fogg, Jonathan Marshall, Tami Kameda, Joshua Solomon, Akshay Rangnekar,
John Boyd, and Bonny Brown. 2001. Web Credibility Research: A Method for
Online Experiments and Early Study Results. (2001).
Designing Trustworthy User Interfaces for the Voluntary Carbon Market To appear in DIS ’22, June 13-June 17, 2021, Virtual Event, USA
[26]
Centre for Eective Altruism. 2020. Introduction to Eective Altruism. https:
//www.eectivealtruism.org/articles/introduction-to-eective-altruism/
[27]
Michael Froehlich, Charlotte Kobiella, Albrecht Schmidt, and Florian Alt. 2021.
Is It Better With Onboarding? Improving First-Time Cryptocurrency App Expe-
riences (DIS ’21). Association for Computing Machinery, New York, NY, USA,
78–89. https://doi.org/10.1145/3461778.3462047
[28]
Michael Froehlich, Maurizio Raphael Wagenhaus, Albrecht Schmidt, and Florian
Alt. 2021. Don’t Stop Me Now! Exploring Challenges Of First-Time Cryptocur-
rency Users. In Designing Interactive Systems Conference 2021 (Virtual Event, USA)
(DIS ’21). Association for Computing Machinery, New York, NY, USA, 138–148.
https://doi.org/10.1145/3461778.3462071
[29]
Michael Fröhlich, Felix Gutjahr, and Florian Alt. 2020. Don’t Lose Your Coin!
Investigating Security Practices of Cryptocurrency Users. In Proceedings of the
2020 ACM Designing Interactive Systems Conference (Eindhoven, Netherlands)
(DIS ’20). Association for Computing Machinery, New York, NY, USA, 1751–1763.
https://doi.org/10.1145/3357236.3395535
[30]
Diego Gambetta. 2000. Can We Trust Trust? Trust: Making and Breaking Cooper-
ative Relations (2000), 18. https://www.researchgate.net/publication/255682316
[31]
David Gefen. 2002. Ree ctions on the Dimensions of Trust and Trustworthiness
among Online Consumers. SIGMIS Database 33, 3 (aug 2002), 38–53. https:
//doi.org/10.1145/569905.569910
[32]
David Gefen and Detmar W. Straub. 2004. Consumer trust in B2C e-Commerce
and the importance of social presence: experiments in e-Products and e-Services.
Omega 32, 6 (Dec. 2004), 407–424. https://doi.org/10.1016/j.omega.2004.01.006
[33]
Dennis A. Gioia, Kevin G. Corley, and Aimee L. Hamilton. 2013. Seeking Qual-
itative Rigor in Inductive Research: Notes on the Gioia Methodology. Orga-
nizational Research Methods 16, 1 (Jan. 2013), 15–31. https://doi.org/10.1177/
1094428112452151
[34]
Edward L. Glaeser, David I. Laibson, Jose A. Scheinkman, and Christine L. Soutter.
2000. Measuring Trust*. Quarterly Journal of Economics 115, 3 (Aug. 2000),
811–846. https://doi.org/10.1162/003355300554926
[35]
Duncan Grimson, Stephen Knowles, and Philip Stahlmann-Brown. 2020. How
close to home does charity begin? Applied Economics 52, 34 (July 2020), 3700–3708.
https://doi.org/10.1080/00036846.2020.1720906
[36]
Lars Hallén, Jan Johanson, and Nazeem Seyed-Mohamed. 1991. Interrm Adap-
tation in Business Relationships. (1991).
[37]
Barbara Haya, Danny Cullenward, Aaron L. Strong, Emily Grubert, Robert Heil-
mayr, Deborah A. Sivas, and Michael Wara. 2020. Managing uncertainty in
carbon osets: insights from California’s standardized approach. Climate Policy
20, 9 (Oct. 2020), 1112–1126. https://doi.org/10.1080/14693062.2020.1781035
[38]
Kasper Hornbæk. 2006. Current practice in measuring usability: Challenges to
usability studies and research. International Journal of Human-Computer Studies
64, 2 (Feb. 2006), 79–102. https://doi.org/10.1016/j.ijhcs.2005.06.002
[39]
Peter Howson, Sarah Oakes, Zachary Baynham-Herd, and Jon Swords. 2019.
Cryptocarbon: The promises and pitfalls of forest protection on a blockchain.
Geoforum 100 (2019), 1–9. https://doi.org/10.1016/j.geoforum.2019.02.011
[40] IPCC. 2019. Special Report on the impacts of global warming of 1.5°C.
[41]
Janne Jalava. 2003. From Norms to Trust: The Luhmannian Connections between
Trust and System. European Journal of Social Theory 6, 2 (May 2003), 173–190.
https://doi.org/10.1177/1368431003006002002
[42]
Theodore Jensen, Mohammad MaiHasan Khan, Md Abdullah Al Fahim, and
Yusuf Albayram. 2021. Trust and Anthropomorphism in Tandem: The Interrelated
Nature of Automated Agent Appearance and Reliability in Trustworthiness Percep-
tions. Association for Computing Machinery, New York, NY, USA, 1470–1480.
https://doi.org/10.1145/3461778.3462102
[43]
Dan J. Kim, Donald L. Ferrin, and H. Raghav Rao. 2008. A trust-based consumer
decision-making model in electronic commerce: The role of trust, perceived
risk, and their antecedents. Decision Support Systems 44, 2 (Jan. 2008), 544–564.
https://doi.org/10.1016/j.dss.2007.07.001
[44] Klima-Kollekte. 2021. Osetting. https://klima-kollekte.de/en/osetting
[45]
Bran Knowles. 2016. Emerging Trust Implications of Data-Rich Systems. IEEE
Pervasive Computing 15, 4 (2016), 76–84. https://doi.org/10.1109/MPRV.2016.68
[46]
Bran Knowles, Oliver Bates, and Maria Håkansson. 2018. This Changes Sustainable
HCI. Association for Computing Machinery, New York, NY, USA, 1–12. https:
//doi.org/10.1145/3173574.3174045
[47]
James Konow. 2006. Mixed Feelings: Theories and Evidence of Warm Glow and
Altruism. SSRN Electronic Journal (2006). https://doi.org/10.2139/ssrn.980349
[48]
Grammateia Kotsialou, Karlygash Kuralbayeva, and Tim Laing. 2021. Forest
carbon osets over a smart ledger. Available at SSRN (2021).
[49]
Silvana Krasteva and Huseyin Yildirim. 2013. (Un)Informed charitable giving.
Journal of Public Economics 106 (Oct. 2013), 14–26. https://doi.org/10.1016/j.
jpubeco.2013.06.011
[50]
Louisa Küchler, Guido Hertel, and Meinald T. Thielsch. 2020. Are you willing
to donate?: relationship between perceived website design, trust and donation
decisions online. In Proceedings of the Conference on Mensch und Computer. ACM,
Magdeburg Germany, 223–227. https://doi.org/10.1145/3404983.3409993
[51]
Olli Lagerspetz. 1998. Trust: the tacit demand (1 ed.). Springer Science & Business
Media, Luxemburg.
[52]
Jonathan Lazar. 2017. Research methods in human computer interaction (2nd
edition ed.). Elsevier, Cambridge, MA.
[53]
John D Lee and Katrina A See. 2004. Trust in Automation: Designing for Appro-
priate Reliance. Human Factors (2004).
[54]
Larry Lohmann. 2005. Marketing and making carbon dumps: Commodication,
calculation and counterfactuals in climate change mitigation. Science as Culture
14, 3 (Sept. 2005), 203–235. https://doi.org/10.1080/09505430500216783
[55]
Larry Lohmann. 2006. Carb on trading: A critical conversation on climate change,
privatisation and power. (2006).
[56]
Larry Lohmann. 2008. Carbon Trading, Climate Justice and the Production
of Ignorance: Ten examples. Development 51, 3 (Sept. 2008), 359–365. https:
//doi.org/10.1057/dev.2008.27
[57]
Roger C Mayer and James H Davis. 2021. An Integrative Model of Organizational
Trust. (2021).
[58]
D. Harrison McKnight, Vivek Choudhury, and Charles Kacmar. 2002. De-
veloping and Validating Trust Measures for e-Commerce: An Integrative Ty-
pology. Information Systems Research 13, 3 (Sept. 2002), 334–359. https:
//doi.org/10.1287/isre.13.3.334.81
[59]
Makuochi Nkwo, Rita Orji, and John Ugah. 2018. Mobile persuasion: promot-
ing clean and sustainable environment. In Proceedings of the Second African
Conference for Human Computer Interaction: Thriving Communities. 1–4.
[60]
Makuochi Nkwo, Banuchitra Suruliraj, Rita Orji, and John Ugah. [n.d.]. Socially-
oriented Persuasive Strategies and Sustainable Behavior Change: Implications
for Designing for Environmental Sustainability.
[61]
Donald Arthur Norman. 2004. Emotional Design: Why We Love (or Hate) Every-
day Things.
[62]
Tiago Oliveira, Matilde Alhinho, Paulo Rita, and Gurpreet Dhillon. 2017. Mod-
elling and testing consumer trust dimensions in e-commerce. Computers in Hu-
man Behavior 71 (June 2017), 153–164. https://doi.org/10.1016/j.chb.2017.01.050
[63]
Dhiren Patel, Benita Britto, Sanidhya Sharma, Kaustubh Gaikwad, Yash Dusing,
and Mrinal Gupta. 2020. Carbon Credits on Blockchain. In 2020 International
Conference on Innovative Trends in Information Technology (ICITIIT). 1–5. https:
//doi.org/10.1109/ICITIIT49094.2020.9071536
[64]
Supavich (Fone) Pengnate and Rathindra Sarathy. 2017. An experimental in-
vestigation of the inuence of website emotional design features on trust in
unfamiliar online vendors. Computers in Human Behavior 67 (Feb. 2017), 49–60.
https://doi.org/10.1016/j.chb.2016.10.018
[65]
Charlene Pleger Bebko. 2000. Ser vice intangibility and its impact on consumer
expectations of service quality. Journal of Services Marketing 14, 1 (Feb. 2000),
9–26. https://doi.org/10.1108/08876040010309185
[66]
Robin Pomeroy. 2020. One trillion trees - World Economic Forum launches plan
to help nature and the climate. https://www.weforum.org/agenda/2020/01/one-
trillion-trees- world-economic-forum- launches-plan- to-help- nature-and-the-
climate/
[67]
Aseem Prakash and Mary Kay Gugerty. 2010. Trust but verify? Voluntary reg-
ulation programs in the nonprot sector. Regulation & Governance (Feb. 2010).
https://doi.org/10.1111/j.1748-5991.2009.01067.x
[68]
Carbon retirement limited. 2009. THE EFFICIENCY OF CARBON OFFSET TING
THROUGH THE CLEAN DEVELOPMENT MECHANISM.
[69]
Jens Riegelsberger and M. Angela Sasse. 2002. Trustbuilders and Trustbusters.
In Towards the E-Society, Beat Schmid, Katarina Stanoevska-Slabeva, and Volker
Tschammer (Eds.). Vol. 74. Kluwer Academic Publishers, Boston, 17–30. https:
//doi.org/10.1007/0-306- 47009-8_2 Series Title: IFIP International Federation for
Information Processing.
[70]
Jens Riegelsberger, M.Angela Sasse, and John D.McCarthy. 2003. The researcher’s
dilemma: evaluating trust in computer-mediated communication. International
Journal of Human-Computer Studies 58, 6 (June 2003), 759–781. https://doi.org/
10.1016/S1071-5819(03)00042- 9
[71]
Denise Christine Rieser and Orlando Bernhard. 2016. Measuring Trust: The
Simpler the Better?. In Proceedings of the 2016 CHI Conference Extended Abstracts
on Human Factors in Computing Systems. ACM, San Jose California USA, 2940–
2946. https://doi.org/10.1145/2851581.2892468
[72]
David Robins, Jason Holmes, and Mary Stansbury. 2010. Consumer health in-
formation on the Web: The relationship of visual design and perceptions of
credibility. Journal of the AmericanSo ciety for Information Science and Technology
61, 1 (Jan. 2010), 13–29. https://doi.org/10.1002/asi.21224
[73]
Adrian Sargeant and Stephen Lee. 2004. Donor Trust and Relationship Com-
mitment in the U.K. Charity Sector: The Impact on Behavior. Nonprot and
Voluntary Sector Quarterly 33, 2 (June 2004), 185–202. https://doi.org/10.1177/
0899764004263321
[74]
Mirjam Seckler, Silvia Heinz, Seamus Forde, Alexandre N. Tuch, and Klaus Opwis.
2015. Trust and distrust on the web: User experiences and website characteristics.
Computers in Human Behavior 45 (April 2015), 39–50. https://doi.org/10.1016/j.
chb.2014.11.064
[75]
Venkatesh Shankar, Glen L. Urban, and Fareena Sultan. 2002. Online trust:
a stakeholder perspective, concepts, implications, and future directions. The
Journal of Strategic Information Systems 11, 3-4 (Dec. 2002), 325–344. https:
//doi.org/10.1016/S0963-8687(02)00022- 7
To appear in DIS ’22, June 13-June 17, 2021, Virtual Event, USA Guzij et al.
[76]
Bhiru Shelat and Florian N Egger. 2002. What Makes People Trust Online
Gambling Sites? (2002).
[77]
Blair H Sheppard and Dana M Sherman. 1998. The Grammars of Trust: A Model
and General Implications. (1998).
[78]
Kevin Smith, Oscar Reyes, Timothy Byakola, and Carbon Trade Watch. 2007.
The carbon neutral myth: oset indulgences for your climate sins. Transnational
Institute, Amsterdam. OCLC: 778008109.
[79]
Julianne Stanford, Ellen R Tauber, B J Fogg, and Leslie Marable. 2002. Experts vs.
Online Consumers: A Comparative Credibility Study of Health and Finance Web
Sites. (2002).
[80]
Laerd Statistics. [n.d.]. The ultimate IBM
®
SPSS
®
Statistics guides. https:
//statistics.laerd.com/
[81]
Ulrike Steinbrück, Heike Schaumburg, Sabrina Duda, and Thomas Krüger. 2002.
A Picture Says More Than A Thousand Words - Photographs As Trust Builders
in E-Commerce Websites. (2002).
[82]
M. T.Thielsch, I. Blotenb erg, and R. Jaron.2014. User Evaluation of Websites:From
First Impression to Recommendation. Interacting with Computers 26, 1 (Jan. 2014),
89–102. https://academic.oup.com/iwc/article-lookup/doi/10.1093/iwc/iwt033
[83]
Meinald T. Thielsch and Gerrit Hirschfeld. 2019. Facets of Website Content.
Human–Computer Interaction 34, 4 (July 2019), 279–327. https://doi.org/10.1080/
07370024.2017.1421954
[84]
Lauren Thornton, Bran Knowles, and Gordon Blair. 2021. Fifty Shades of Grey:
In Praise of a Nuanced Approach Towards Trustworthy Design. In Proceedings of
the 2021 ACM Conference on Fairness, Accountability, and Transparency (Virtual
Event, Canada) (FAccT ’21). Association for Computing Machinery, New York,
NY, USA, 64–76. https://doi.org/10.1145/3442188.3445871
[85]
Ehsan Toreini, Mhairi Aitken, Kovila Coopamootoo, Karen Elliott, Carlos Gon-
zalez Zelaya, and Aad van Moorsel. 2020. The Relationship between Trust
in AI and Trustworthy Machine Learning Technologies. In Proceedings of the
2020 Conference on Fairness, Accountability, and Transparency (Barcelona, Spain)
(FAT* ’20). Association for Computing Machinery, New York, NY, USA, 272–283.
https://doi.org/10.1145/3351095.3372834
[86]
Ludwig Trotter, Mike Harding, Chris Elsden, Nigel Davies, and Chris Speed.
2020. A Mobile Platform for Event-Driven Donations Using Smart Contracts.
In Proceedings of the 21st International Workshop on Mobile Computing Systems
and Applications (Austin, TX, USA) (HotMobile ’20). Association for Computing
Machinery, New York, NY, USA, 108. https://doi.org/10.1145/3376897.3379161
[87]
Ludwig Trotter, Mike Harding, Peter Shaw, Nigel Davies, Chris Elsden, Chris
Speed, John Vines, Aydin Abadi, and Josh Hallwright. 2020. Smart Donations:
Event-Driven Conditional Donations Using Smart Contracts On The Blockchain.
In 32nd Australian Conference on Human-Computer Interaction (Sydney, NSW,
Australia) (OzCHI ’20). Association for Computing Machinery, New York, N Y,
USA, 546–557. https://doi.org/10.1145/3441000.3441014
[88]
Lili Wang and Elizabeth Graddy. 2008. Social Capital, Volunteering, and Charita-
ble Giving. VOLUNTAS: International Journal of Voluntary and Nonprot Organi-
zations 19, 1 (March 2008), 23–42. https://doi.org/10.1007/s11266- 008-9055-y
[89]
Ye Diana Wang and Henry H. Emurian. 2005. An overview of online trust:
Concepts, elements, and implications. Computers in Human Behavior 21, 1 (Jan.
2005), 105–125. https://doi.org/10.1016/j.chb.2003.11.008
[90]
Peter Worthy, Ben Matthews, and Stephen Viller. 2016. Trust Me: Doubts and
Concerns Living with the Internet of Things. In Proceedings of the 2016 ACM
Conference on Designing Interactive Systems (Brisbane, QLD, Australia) (DIS ’16).
Association for Computing Machinery, New York, NY, USA, 427–434. https:
//doi.org/10.1145/2901790.2901890
[91]
Cherrie Yang and Deryl Northcott. 2019. How can the public trust charities? The
role of performance accountability reporting. Accounting & Finance 59, 3 (Sept.
2019), 1681–1707. https://doi.org/10.1111/ac.12475
[92]
Akbar Zaheer. 1995. Relational Governance as an Interorganizational Strategy: An
Empirical Test of the Role of Trust in Economic Exchange. Strategic Management
Journal 16, 5 (1995). https://w ww.jstor.org/stable/2486708
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
In this forum we highlight innovative thought, design, and research in the area of interaction design and sustainability, illustrating the diversity of approaches across HCI communities. --- Roy Bendor, Editor
Conference Paper
Full-text available
Engaging first-time users of mobile apps is challenging. Onboarding task flows are designed to minimize the drop out of users. To this point, there is little scientific insight into how to design these task flows. We explore this question with a specific focus on financial applications, which pose a particularly high hurdle and require significant trust. We address this question by combining two approaches. We first conducted semi-structured interviews (n=16) exploring users' meaning-making when engaging with new mobile applications in general. We then prototyped and evaluated onboarding task flows (n=16) for two mobile cryptocurrency apps using the minimalist instruction framework. Our results suggest that well-designed onboarding processes can improve the perceived usability of first-time users for feature-rich mobile apps. We discuss how the expectations users voiced during the interview study can be met by applying instructional design principles and reason that the minimalist instruction framework for mobile onboarding insights presents itself as a useful design method for practitioners to develop onboarding processes and also identify when not to.
Conference Paper
Full-text available
Cryptocurrencies have increasingly gained interest in practice and research alike. Current research in the HCI community predominantly focuses on understanding the behavior of existing cryptocurrency users. Little attention has been given to early users and the challenges they encounter. However, understanding how interfaces of cryptocurrency systems support, impede, or even prevent adoption through new users is essential to develop better, more inclusive solutions. To close this gap, we conducted a user study (n=34) exploring challenges first-time cryptocurrency users face. Our analysis reveals that even popular wallets are not designed for novice users' needs, stopping them when they would be ready to engage with the technology. We identify multiple challenges ranging from general user interface issues to finance and cryptocurrency-specific ones. We argue that these challenges can and should be addressed by the HCI community and present implications for building better cryptocurrency systems for novice users.