There's a growing interest in optimizing User Interface (UI) design to better align with organizational strategies and facilitate improved guidance of users towards specific options in a choice architecture, a kind of user behavior influencing known as "digital nudging". The paper reports on a Design Science Research project introducing a domain-specific modeling method to integrate digital nudging strategies with User Experience (UX) and feature portfolio design. This paper proposes Semantic Digital Nudging as a modeling method that semantically enriches Digital Nudges implemented in UX workflows by linking to target product features. This is deployed as a visual tool that embeds properties which can be queried to assess the friction certain nudges cause on User Experience and to maintain an inventory of by-design nudges. Linking the procedural knowledge in UI usage flows with product features and the nudging patterns makes this a multi-perspective domain-specific modeling approach that can help businesses to plan the layout of user choices, or help the user make more sustainable choices by increasing salience of healthier alternatives.
objectives targeted by them. The portfolio takes the form of feature decomposition graphs, which
can later represent stakeholders goals by analogy. To treat this, our research proposes as artifact
a modeling method expanding from pre-existing modeling languages. Firstly, we add modeling of
Digital Nudging associated to UX flow modeling we developed in previous work; secondly we link
this to an Extended Feature Model. They are interlinked by hyper-references in a demonstrator
that is a modeling tool built inside the ADOxx [
A controversial nudge for UX is friction, which is temporarily delaying acting upon a decision.
This barrier adds steps in-between user actions giving more time to reconsider their choice.
Proposed modeling language abstracts the friction nudge dichotomously in respect to the
definition given in [
The remainder of the article is organized as follows: Section 2 presents an overview of the research methodology. Section 3 discusses related research. Section 4 formulates the problem tackled and puts it into context. Section 5 presents the objectives of the problem treatment and specific requirements. Section 6 exposes design and development decisions for the treatment and Section 7 shows demonstration scenarios. Conclusions end the paper.
For carrying out present research we have aligned our process to the Design Science Research
methodology [
Problem identification and motivation commenced this iteration and is reported in Section 4. A literature review condensed in Section 3 aided in problem and objective definition. We tackle lack of a centralized method for UX designers to represent in a linked fashion diagrammatic designs of UX workflow, Digital Nudging strategies and Feature portfolios. A modeling methodbased treatment enables semantically linking of HCI steps to specific descriptions of digital nudges that reference particular features/objectives. Choice architecture is the design of how options are presented. Proposed treatment can also raise awareness of healthier alternatives since online choice architecture can be cumbersome.
Next, Objectives of a solution were formulated and synthesized in Section 5. UX designers with managers will put the vision into knowledge form: to strategically design User Experiences that realize Digital Nudging policies. Simulations based on quantitative queries should be enabled – i.e., totalize extra time the user takes based on friction nudges and UX design.
Design and development were concerned with defining and implementing functionalities of the
modeling method. These are discussed in large in Section 6. We embedded inside this DSR step
the Agile Modeling Method Engineering (AMME) methodology [
During Demonstration phase we used the modeling tool deployed using ADOxx to model an online flight booking service. A walk through exemplary models and how they reflect Semantic Digital Nudging is showcased in Section 7. At each stage of the iterative DSR process we came back to the method and adjusted design and implementation of the method and models.
As the current DSR iteration is not finalized, a thorough Evaluation has not yet been made. Instead, the focus was on demonstrating implementation feasibility by requirements coverage of exemplary models. The final step of Communication is instantiated in the present report.
To define any aspect of a choice architecture that changes people’s cognition and behavior [
Works such as [
To better fit client needs, the authors of [17] designed an architecture for smart digital nudging-based recommender systems. Gently pushing public transportation users towards choosing greener alternatives involves collection of integrated data, processing and creation of user profiles. This would mitigate the user annoying that added friction causes by being aware to user choices and reducing recommendations when the nudging strategy fails, instead of feeding them the same UX . The idea of digital nudging for guiding a more sustainable usage of work, health and commerce apps is captured in [18]. The framework combines smart feedback and reminders with defaulting for the good of the user.
The work of [19] deploys a digital nudge enforcing tool that monitors code debt in Credit Suisse software. It provides visual cues to address debt responsibility at its roots – in the decisions of each developer. A method for digital nudging of recommendations in e-commerce by the authors of [20] uses a Knowledge Graph to feed the engine business knowledge besides traditional customer related data. This is close to the present work, by using AMME to setup a diagrammatic tool for a business-centric view of prioritization rules. Models are exported in RDF form to power knowledge-intensive nudging in a recommender engine.
The authors of [
We decompose features targeted by nudges in Feature Diagrams – first introduced in 1990 by the Feature-Oriented Domain Analysis (FODA) method [24]. The semantics of Feature Models have been analyzed and formalized by the authors of [25] following implementation variations in software product lines. We also based our metamodel on their analysis of the Extended Feature Model in [26]. Its increased flexibility stems from using UML-like cardinalities over the base FODA model and moving optionality from feature instance to the decomposition relation.
The field of HCI is dominated by behavior-oriented empirical studies, lacking in design theories, methods and artifacts (except for some non-scientific practitioner recipes). We felt that reviewed literature was too strict, with works either reviewing nudging patterns and scenarios, others focused on behavior changing. Few tackled real data-enhanced methods of smart nudging and enterprises could benefit from an encompassing method of strategic HCI design. A cause of this gap is heterogenous management and representation of UX, product features and nudging knowledge. We fill this gap by introducing a method for UX designers and marketing planners to represent in a linked way knowledge that is typically represented disparately. In this proposal, feature portfolios and digital nudges that strategically target the items should be attached to each step in the UX workflow. We are repurposing workflow patterns to understand the phenomenon of digital nudging, to systematize its design patterns, and ultimately to offer a domain-specific modeling method for incorporating such patterns into UX workflows and for assessing/simulating their impact on UX.
Design Science Research addresses artifacts in context [
We formulate the problem statement in the tradition of the DSR template [27]: Improve nudging knowledge management (problem context) ...by treating it with an integrative agile modeling method for Semantic Digital Nudging (artifact)
...to satisfy a need for aligning UX workflows with nudging mechanisms and productservice features (requirements)
...in order to facilitate the reuse, analytics and management of nudges aligning better to organizational strategies and user satisfaction (goals)
The main objective is to support visual design of methods to persuade users towards specific options as they are guided through the UX. Secondly, the resulting conceptual models should enable analytics of Digital Nudging policies efficiency while considering impact on UX. This requires embedding values from post-deployment interviews with users assessing their effort of avoiding certain nudges. Thirdly, for competency-based evaluation and finer capturing of model semantics, a sufficient amount of unpacking has to be iteratively made for concepts and relations in the metamodels. Lastly, the resulting modeling method should be agile in response to requirements that might arise and support an iterative process. This is covered by adhering to the AMME methodology.
From the exemplary scenario of flight booking described during problem identification the solution addresses the following requirements: • To model with added semantics the UX workflow – i.e., highlighting economic paths for user and actions caused by nudges (UX designer) • To model in the same diagram Semantic Digital Nudges linked to specific steps in UX (Business owner decides where to put these nudges) • To model using feature-based decomposition an Extended Feature Diagram aligned to the ontology established by [26] (Managers) • To set hyperlinks between nudge implementation instances in UX/Digital Nudging diagram and targeted features in Extended Feature Diagram (UX designer) • Clear visual notation for enhancing understanding of diagrams by people with various technical skill levels • To enable analytics of UX impact and nudging efficiency through concept instance attributes – i.e., containing user feedback encoded as integers • To allow queries at a sufficient level of detail for Knowledge Management that facilitate better (re)design of models and propel Digital Transformation
The requirement of assessing friction nudges impact on UX partly answers concerns of [
To integrate specificities of Digital Nudging, UX and Feature Models we conceived as artifact a
domain-specific conceptual modeling method. At this stage of the DSR cycle we covered
eponymous steps from the lifecycle of the methodology Agile Modeling Method Engineering
(AMME). AMME hybrid modeling framework enables developing the modeling language,
procedure and algorithms [
First iteration starts with the create step by acquiring knowledge and eliciting requirements for modeling Digital Nudging. Preliminary requirements and conceptual knowledge were drawn in a discussion. We agreed to start by extending a modeling method that we have previously tailored for representing RPA/UX. We filtered out concepts pertaining to automation and preserved the hardNext relation that denotes succession between two process steps. We created a preliminary model draft on the UX of booking a flight from a low-cost company. It included notes as concept placeholders, arrows and legends which progressively responded to the solution requirements. It served as an agile-proof pre-metamodel through the lifecycle.
Additional requirements arose and we returned to the create step. We extended the modelling
method with a Digital Nudge concept and the implements relation that links a UX Action to a
Digital Nudge. We conceptualized the ten digital nudges from the taxonomy of [
In the design step we created a metamodel for Digital Nudging in the context of UX. It is depicted in Figure 1. Concepts under the WorkflowElement category existed in the RPA/UX metamodel of the previous project. These can follow one another (through followedBy relation) and are either containers (Activity) or atomic steps such as Action, Decision or Start and End States which replicate BPMN [30] events. An Action is either See which is the user identifying a particular UI element (substitutable by Computer Vision) or an Input. In this DSR cycle we have integrated the BoundaryEvent concept from the BPMN standard making it a kind of State that can be attached on the border of Activity instances or stand alone and reference a couple of them. We added a machine-readable property to mark a “Nudge-called action” introduced by friction nudge and semantically differentiate these in queries from regular UX Actions. We have added a new Action subtype, Scroll, which is called upon the user by the usage of nudging.
We have added the Digital Nudge sub-hierarchy in the right part of the figure. A Digital Nudge
can employ several patterns from the ten in [
We started the second and last AMME iteration by extracting knowledge regarding feature models. We based a metamodel depicted in Figure 2 on that of the Extended Feature Diagram (EFD) [26] and deployed a modeling tool. To validate it we modeled decomposition of features in graph form for the flight booking scenario. We then returned to propagate these changes in the Digital Nudging model type. Ultimately, we linked features to Digital Nudge instances by means of an INTERREF attribute which accepts an instance of type Feature from an EFD. Updates were deployed to the exemplary models demonstrated in the upcoming section.
Implementation-wise, we carefully weigh on which concept to apply unpacking and repacking. We implied that when modeling friction nudges there is a delay for the user processing it, and Delay will not propagate into a separate Action type. This is a case of repacking to declutter the diagram. We unpacked the inherited Action by adding the “Nudge-caused action” attribute and Scroll subtype to the pragmatic Input actions. A See Action that implements Nudge or its outgoing causes relation Action target imply a nudge-induced Delay.
A snapshot of the flight ticket booking process for a low-cost company is depicted in Figure 3. Our goal was to book a single passenger ticket fast with minimum cost (paying just for basic features). The options picked in previous screens were consistent with this usage strategy. Embedded UI logic presents in response three advertisements where they try to persuade the user in upgrading to costlier services. Semantic priming presents as stimulus the feature of coverage for “travel uncertainties” in the top grid for the eye to catch it first. This is reinforcing the company’s premium package. Extended features listed might be evaluated by the user’s cognitive system as superior to their status-quo option, which is the cheapest bundle. An interaction which couldn’t be captured in rendered mockup is dynamic addition or subtraction to/from the interface. Appearance of the Privilege Pass grid right after completing Passenger information lures towards exploring the option by disclosing discounts to the overall ticket price.
We translated the UX workflow in the Passengers screen using the Semantic Digital Nudging modeling method. It is depicted in Figure 4. We will demonstrate the Digital Nudges implemented by Actions in the “Complete PASSENGERS Screen” Activity. This directed relation is visually denoted by the implements text. It denotes intention of the choice architect to hint the user into opting for a feature in the organization’s catalogue. The two costlier bundle alternatives, Free flight change and Convenient Traveling, are nudged at extreme positions in the UX process. The flight change option is the first alternative with which the user is overtly persuaded after being dismissed in favor of the Light traveling bundle on a prior screen. The nudge pattern is Reinforcement with the mechanism of persistent/frequent exposure. Both icons that correspond to a hybrid overt-covert pattern are dynamically loaded at the top-left of the concept.
In the case of nudging towards the Privilege Pass, Added Friction delays the user by having to see and process the nudge’s visual presentation. Although this classifies as a nudge-called action we agree not to depict nudge-called See Actions as seeing the nudge is implied. Digital Nudges can cause the user to make an input, i.e., scrolling over the grid advertisement to avoid Convenient Traveling nudge. If they close and/or open novel UI elements, then Digital Nudges morph the UI and symbols decorate the connector accordingly.
Fractionally depicted in Figure 5 is a decomposition hierarchy of features. They are partly targeted by the Digital Nudge instances in associated model. Flight ticket is the root Feature that gets decomposed either mandatory or optionally into sub-features. For that, it is called a Concept. In FODA, the Concept was restricted to a single node and it was the root of the feature tree. To fit the graph structure of model diagrams, we permitted multiple nodes of type Concept as either root or they can exist outside of composition. Credit wallet Concept is an example of stand-alone feature that is the target of a Nudge. It is not part of the flight ticket, though, having a lifetime/scope of its own. Similarly, Privilege Pass subscription is a stand-alone entity, but this time it can become a part of Flight ticket options.
The Multiplicity node specifies the minimum and maximum number of required options from a group of sub-features. Its omission equivalates to AND decomposition with the cardinality being the total number of children. For instance, Light Traveling feature mandatorily consists of Carry bag and Prior day check-in. A Bundle must consist of one of the three options and at most one, and this is an example of XOR decomposition.
UX Activities can have Boundary Events attached to them as seen in the left part of Figure 6. They are exceptional events borrowed from the BPMN standard that happen conditionally or after certain time passes. Here the designer uses one on the last screen of payment to hurry the user up if a time threshold is reached without them advancing through the booking. Since prices might fluctuate, this induces the idea of price and ticket scarcity based on the hypthosesis formulated by the authors of [31]. It states that products with time-related purchase pressure cues will be chosen more often.
Exemplified Scarcity nudge can be edited to link the targeted feature from an EFD, or by adding a policy explanation in the machine-readable description property. Editing screen appears on the right side of Figure 6. Because this nudge requires the user to click a button it adds Friction to the UX, which can be quantitatively assessed from end-users. The collected values can be retrieved later by queries that assess an estimated delay in seconds added by the Digital Nudge.
The paper reports on a Design Science project where an artifact in the form of an agile modeling
method was engineered. It enables Semantic Digital Nudging as a strategy for integrated design
of Digital Nudges at the level of User Experience and with semantic enrichment from Extended
Feature Diagrams. The goal of the artifact is to support the process of creating smarter nudges
that better integrate with product feature strategy and are enhanced by the UI/UX design. This is
achieved by better aligning UX requirements to nudging policies, product feature portfolio and
user satisfaction since the two model types are understandable at a high level by less-technical
people. The inclusion of machine-readable attributes that asses level of friction to UX helps digital
choice architects achieve their organizational goals by understanding UX and actual nudging
effects on users. As per [32] designers must be aware of the ethical implications of nudges, but
this is a topic beyond the scope of this paper (see [33] for a discussion on nudging ethics).
Therefore, we align our goals to those of [
We intend to end the iteration with a competency-question based evaluation upon collected
requirements. Questions will cover post deployment analysis on UX by using attributes on
modeled instances that trace friction nudges. Cross-model Knowledge extraction will be assessed.
In future DSR iterations there remains to investigate additional nudging patterns in literature.
Mainly the set of Digital Dark Nudges in the work of [34] and the taxonomical differences of the
23 nudges classified by the authors of [
The presented work has received financial support through the project: Integrated system for automating business processes using artificial intelligence, POC/163/1/3/121075 - a Project Cofinanced by the European Regional Development Fund (ERDF) through the Competitiveness Operational Programme 2014-2020. CHI Conference on Human Factors in Computing Systems, Association for Computing Machinery, New York, NY, USA, 2019: pp. 1–15. https://doi.org/10.1145/3290605.3300733. [16] C. Meske, T. Potthoff, The DINU-Model – A Process Model for the Design of Nudges, Researchin-Progress Papers. (2017) 2587–2597. https://aisel.aisnet.org/ecis2017_rip/11. [17] R. Karlsen, A. Andersen, Recommendations with a Nudge, Technologies. 7 (2019) 45.
https://doi.org/10.3390/technologies7020045. [18] M. Sobolev, Digital Nudging: Using Technology to Nudge for Good, (2021).
https://doi.org/10.2139/ssrn.3889831. [19] K. Haki, A. Rieder, L. Buchmann, A. W.Schneider, Digital nudging for technical debt management at Credit Suisse, European Journal of Information Systems. 32 (2023) 64–80. https://doi.org/10.1080/0960085X.2022.2088413. [20] D.-A. Sitar-Tăut, D. Mican, R.A. Buchmann, A knowledge-driven digital nudging approach to recommender systems built on a modified Onicescu method, Expert Systems with Applications. 181 (2021) 26. https://doi.org/10.1016/j.eswa.2021.115170. [21] B.M. Berens, H. Dietmann, C. Krisam, O. Kulyk, M. Volkamer, Cookie Disclaimers: Impact of Design and Users’ Attitude, in: Proceedings of the 17th International Conference on Availability, Reliability and Security, Association for Computing Machinery, New York, NY, USA, 2022: pp. 1–20. https://doi.org/10.1145/3538969.3539008. [22] C. Meske, I. Amojo, P. Mohr, Digital Nudging to Increase Usage of Charity Features on ECommerce Platforms, in: Proceedings of the 15th International Conference of Business Informatics 2020., Potsdam, Germany, 2020: pp. 1203–1218. https://doi.org/10.30844/wi_2020_k5-meske. [23] S. Lu, G. (Gordon) Chen, K. Wang, Overt or covert? Effect of different digital nudging on consumers’ customization choices, NBRI. 12 (2021) 56–74. https://doi.org/10.1108/NBRI12-2019-0073. [24] K.C. Kang, S.G. Cohen, J.A. Hess, W.E. Novak, A.S. Peterson, Feature-Oriented Domain Analysis (FODA) Feasibility Study, Defense Technical Information Center, Fort Belvoir, VA, 1990. https://doi.org/10.21236/ADA235785. [25] P.-Y. Schobbens, P. Heymans, J.-C. Trigaux, Y. Bontemps, Generic semantics of feature diagrams, Computer Networks. 51 (2007) 456–479. https://doi.org/10.1016/j.comnet.2006.08.008. [26] M. Riebisch, K. Böllert, D. Streitferdt, I. Philippow, Extending Feature Diagrams with UML Multiplicities, in: Sixth Conference on Integrated Design and Process Technology (IDPT2002), Pasadena, CA, 2002. [27] R.J. Wieringa, Design Science Methodology Lecture (192320820) Slides, (2015). https://wwwhome.ewi.utwente.nl/~roelw/DSMSlidesBinder2016.pdf (accessed June 21, 2023). [28] R.H. Thaler, Nudge, not sludge, Science. 361 (2018) 431–431.
https://doi.org/10.1126/science.aau9241. [29] R.A. Buchmann, D. Karagiannis, Agile Modelling Method Engineering: Lessons Learned in the ComVantage Research Project, in: J. Ralyté, S. España, Ó. Pastor (Eds.), The Practice of Enterprise Modeling, Springer International Publishing, Cham, 2015: pp. 356–373. https://doi.org/10.1007/978-3-319-25897-3_23. [30] BPMN Specification - Business Process Model and Notation, (n.d.). https://www.bpmn.org/ (accessed August 5, 2023). [31] D. Djurica, K. Figl, The Effect of Digital Nudging Techniques on Customers’ Product Choice and Attitudes towards E-Commerce Sites, in: HUMAN-COMPUTER INTERACTION (SIGHCI), Association for Information Systems (AIS) eLibrary, Boston, 2017: pp. 1–7. https://aisel.aisnet.org/amcis2017/HumanCI/Presentations/13. [32] M. Weinmann, C. Schneider, J. vom Brocke, Digital Nudging, Bus Inf Syst Eng. 58 (2016) 433– 436. https://doi.org/10.1007/s12599-016-0453-1. [33] C.R. Sunstein, Nudging and Choice Architecture: Ethical Considerations, (2015).
https://papers.ssrn.com/abstract=2551264 (accessed August 5, 2023). [34] F.J. Costello, J. Yun, K.C. Lee, Digital Dark Nudge: An Exploration of When Digital Nudges Unethically Depart, in: Proceedings of the 55th Hawaii International Conference on System Sciences, 2022. https://doi.org/10.24251/HICSS.2022.531.