=Paper=
{{Paper
|id=Vol-1953/healthRecSys17_paper_4
|storemode=property
|title=Towards Argumentation-based Recommendations for Personalised Patient Empowerment
|pdfUrl=https://ceur-ws.org/Vol-1953/healthRecSys17_paper_4.pdf
|volume=Vol-1953
|authors=Juan Manuel Fernandez,Marco Mamei,Stefano Mariani,Felip Miralles,Alexander Steblin,Eloisa Vargiu,Franco Zambonelli
|dblpUrl=https://dblp.org/rec/conf/recsys/FernandezM0MSVZ17
}}
==Towards Argumentation-based Recommendations for Personalised Patient Empowerment==
https://ceur-ws.org/Vol-1953/healthRecSys17_paper_4.pdf
Towards Argumentation-based Recommendations
for Personalised Patient Empowerment
Juan Manuel Fernández Marco Mamei
Felip Miralles Stefano Mariani
Alexander Steblin Franco Zambonelli
Eloisa Vargiu DISMI—Università di Modena e Reggio Emilia
Eurecat Technology Center, eHealth Unit Reggio Emilia, Italy
Barcelona, Spain name.surname@unimore.it
name.surname@eurecat.org
ABSTRACT [1]. Although the idea of patient empowerment was introduced
Patient empowerment is a key issue in healthcare. Approaches to in- to healthcare in the 1970s [20], its popularity emerged in the mid
crease patient empowerment encompass patient self-management 1990s [18], and became a feasible reality only in 2000s thanks to
programs. In this paper we present ArgoRec, a recommender sys- the advent of Web 2.0 [5]. In general, strategies to increase pa-
tem that exploits argumentation for leveraging explanatory power tient empowerment address two aspects of patients’ experience
and natural language interactions so as to improve patients’ user [19]: (i) disease management and (ii) relationships with healthcare
experience and quality of recommendations. ArgoRec is part of a providers. Approaches to increase patient empowerment vary from
great effort concerned with supporting complex chronic patients in, patient self-management programs [16], to promoting patient in-
for instance, their daily life activities after hospitalisation, pursued volvement in treatment decision-making [13], to facilitating the
within the CONNECARE project by following a co-design approach physician-patient interaction [17].
to define a comprehensive Self-Management System. In this paper, we present a recommender system named ArgoRec,
which is part of the great effort for providing support to complex
CCS CONCEPTS chronic patients pursued by the CONNECARE project [23], by fol-
lowing a co-design approach. ArgoRec distinctive feature is that it
• Applied computing → Consumer health; Health care in-
relies on argumentation to provide justifiable and personalised rec-
formation systems; • Information systems → Expert systems;
ommendations, increasing accuracy of recommendations based on
• Human-centered computing → Ubiquitous and mobile comput-
continuously monitored data while improving patients’ user expe-
ing systems and tools; • Computing methodologies → Discourse,
rience. Current commercial solutions, in fact, aim to keep patients
dialogue and pragmatics; Multi-agent systems;
autonomous by providing them with wearable, non-intrusive de-
vices (e.g., wristbands and medical devices), paired with proprietary
KEYWORDS
smartphone apps. Nevertheless, recommendations are predefined
Patient empowerment; Personalised care; Recommendation; Argu- by providers and usually cannot be personalised. On the contrary,
mentation; Connecare; ArgoRec clinicians interested in monitoring activities, health-status, and pos-
ACM Reference format: sibly habits, prefer to set the goals to be achieved by each patient
Juan Manuel Fernández, Felip Miralles, Alexander Steblin, Eloisa Vargiu, (e.g., number of steps per day) and expect the system to generate
Marco Mamei, Stefano Mariani, and Franco Zambonelli. Towards Argumentation- tailored recommendations accordingly—as ArgoRec does.
based Recommendations for Personalised Patient Empowerment. In Pro- The rest of the paper is organised as follows. In Section 2 a
ceedings of the Second International Workshop on Health Recommender Sys- minimal and necessary background on argumentation is given.
tems co-located with ACM RecSys 2017, Como, Italy, August 2017 (RecSys’17) ,
Section 3 presents the proposed model and its architectural design.
4 pages.
Section 4 discusses the benefits and challenges stemming from
preliminary results. Section 5 ends the paper with final remarks
1 INTRODUCTION and future directions.
Patient empowerment is a key issue in current healthcare that
should be seen as both an individual and a community process.
Four components are fundamental to the process: (i) understand- 2 ON ARGUMENTATION
ing by the patient of her/his role; (ii) acquisition by patients of Argumentation is amongst the most natural ways people interact
sufficient knowledge to be able to engage with their healthcare through dialogue [22]: people argue by making claims, attack oth-
provider(s); (iii) patient skills; and (iv) a facilitating environment ers’ ones, and provide further premises for supporting own ones,
with the goal of winning a debate. Computational argumentation is
a research thread concerned with designing computational mod-
HealthRecSys’17, August 2017, Como, Italy
© 2017. Copyright for the individual papers remains with the authors. Copying per- els and algorithms to analyse and construct arguments and their
mitted for private and academic purposes. This volume is published and copyrighted relationships with the aim of enabling automatic reasoning over
by its editors. acceptability of arguments [15].
HealthRecSys’17, August 2017, Como, Italy J.M. Fernández et al.
In abstract argumentation arguments are considered as atomic recommendation the message to dispatch to the patient for
units and the only considered relation is the attack one, meaning engagement, reward, or warning, depending on her/his ad-
arguments are in conflict [8], whereas in structured argumentation herence, or the one to be sent to the clinician for continuous
arguments may be constituted by claims (“what to be proven true”) follow-up (in this case, it’s called feedback). According to
and premises (“what helps proving something true”), and relations the corresponding adherence, recommendations may have
amongst them also encompass the support one, linking premises to a punctuation from 1 (“very bad”) to 5 (“very good”), thus
claims [4]. Moreover, attack relations are further divided into rebut- messages sent accordingly: an alert for low punctuation (e.g.,
tal, in case two claims clash, and undercut, when a claim contrasts “You’ve to be more active. Go out and take a walk!”) and a
the premise of the attacked claim. reward for a high one (e.g., “Wonderful! Walk 100 steps more
Many different argumentation frameworks exist, extending the and you’ll reach the goal!”).
notion of argument or relation, or both. For instance, weighted strategy the criteria guiding decision making about how to
[9] and value-based [3] frameworks attach quantitative labels to compute the adherence, and which recommendation/feedback
relations to express, respectively, strength of arguments over others. to send, when.
These kind of schemes are especially useful in those open and highly recommendation engine the component responsible of gen-
dynamic scenarios in which the relevance of arguments is likely to erating and dispatching recommendations and feedbacks,
change over time, i.e., due to acquisition of new information. based on the patients’ adherence regarding their fulfillment
In this paper, we exploit argumentation for (i) empowering rec- of prescriptions, and on a dynamically configurable strategy.
ommendation systems with explanatory power regarding why and In ArgoRec, recommendations and feedback are interpreted as
how recommendations are provided, and (ii) improve patients’ user arguments, whose claims (i.e. the fact that the patient is doing well
experience through natural language interactions—as discussed in or not) are supported by premises constituted by the patient’s ad-
Section 3. In particular, we adopt the simple structured argumen- herence. The strength of support relations is dynamically computed
tation framework depicted in Figure 1 as an argumentation graph, (and adjusted), and depends on the time window that the adherence
where darker nodes are claims whereas lighter ones are premises of the patient refers to: recent activity events (that is, fulfillment
and shaded boxes are whole arguments, solid arrows are attack to more recent prescriptions) are stronger premises with respect
relations whereas dashed ones are support ones – darker ones are to more ancient events. Accordingly, attack relations between ar-
rebuttals and lighter ones are undercuts –, and the thickness of guments are possible because the recommendation engine may be
lines represents the strength of the relation. This serves well the tempted to generate conflicting recommendations based on differ-
purpose of discussing the benefits and challenges of argumentation ent time windows, i.e., focusing on the adherence level (memoryless)
based recommendations (Section 4), while keeping the paper acces- versus the adherence profile (historical). In this case, argumentation
sible to readers unfamiliar with process algebraic descriptions of helps ArgoRec to generate the most correct recommendation (or
argumentation frameworks’ semantics. feedback), by exploiting argumentation-based reasoning to select
Although the idea of using argumentation to improve recom- the stronger claim—that is, the one supported by the strongest
mendations is not novel [2, 6], to the best of our knowledge this is premises. Figure 1 depicts an example argumentation graph in
the first attempt to exploit it in healthcare. which recommendation “keep going” is the strongest argument,
thus gets generated and dispatched. Essentially, despite compari-
3 SYSTEM MODEL & ARCHITECTURE son of latest fulfillment event (f ul f illmenti,t ) with previous one
This section presents our Argumentation-based Recommender (f ul f illmenti,t −1 ) suggests to warn the patient about the need for
system, ArgoRec, by first describing its model & inner functioning improvement (recommendation “must improve”) – since her/his
(Subsection 3.1), and then discussing the architecture of the overall adherence is worsening –, the fact that there is still time left to
self-management ecosystem it is part of (Subsection 3.2). complete prescription (prescriptioni ) steers arguments’ strength in
favour of recommendation “keep going”, to further motivate the
patient.
3.1 System Model Besides correctness, this way ArgoRec can, on the one hand,
ArgoRec revolves around the following main abstractions: provide to patients more convincing recommendation messages, by
prescription any kind of prescription made by a clinician to a
given patient to monitor, e.g., physical activities, health sta- halfway
tus through medical devices and/or suitable questionnaires, adherencei
prescriptioni time lefti > 0
taking medications, and so on.
adherence the adherence of the patient to the clinician’s pre- fulfillmenti,t
scriptions, both regarding individual prescriptions (adher-
“keep going”
ence level) and their history based on a given time window
(adherence profile). fulfillmenti,t-1
worsening
fulfillment the fulfillment of a prescription achieved by the adherencet “must improve”
patient, necessary to measure the patient’s adherence—either
automatically (e.g., through an activity tracker) or manually
(e.g., by the patient her/him-self tracking taken tablets). Figure 1: Example of argumentation graph exploited by ArgoRec.
Argumentation-based Recommendations for Personalised Patient Empowerment HealthRecSys’17, August 2017, Como, Italy
motivating and explaining the reasons behind them (the why) and,
on the other hand, provide to clinicians insights on the decision
making process leading to that precise feedback (the how). Both can
be achieved by navigating the argumentation (sub)graph whose
claim is the recommendation or feedback itself to, for instance, gen-
erate explanation sentences through Natural Language Processing
(NLP) techniques and argumentation mining—as better discussed
in Subsection 4.1.
To deliver its functionalities, ArgoRec works as follows (see
also Figure 2). Whenever an activity fullfillment event is received:
(i) it is checked against the corresponding prescription to compute
adherence level of the patient and update her/his adherence profile,
depending on the configured strategy (i.e. defining how to weight
older vs. newer events); (ii) new arguments are generated accord-
ingly and added to ArgoRec argumentation graph (i.e., an “halfway”
fullfillment may support a “keep going” recommendation); and (iii)
weights of relations are updated depending on the newly-added ar- Figure 2: Flow of data regarding the prescription of a physical ac-
tivity.
guments (i.e. new premises for a claim increasing support strength)
and ArgoRec’s own strategy (i.e. decreasing strength of arguments
as time flows). Finally, periodically and depending on the config- short-term changes in patients behaviour. This will serve as a first
ured policies, ArgoRec generates recommendations and feedback indication of whether argumentation helps motivating patients.
based on the strongest argument(s) in the graph—i.e. navigating In this Section we briefly summarise the key benefits that we en-
the graph to generate sentences through NLP. visage in using the proposed recommender system (Subsection 4.1)
as well as the challenges to be faced by the SMS and ArgoRec for
3.2 System Architecture deployment in production (Subsection 4.2)1 .
ArgoRec is part of a Self-Management System (SMS) developed
within the CONNECARE project and aimed at monitoring patients 4.1 Key Benefits
habits in terms of physical activities, health status, taking medi- Argumentation may play a crucial role in dealing with the fear of
cations, as well as nutrition. It consists of, among others, an app algocracy [7], that is, of having our everyday life influenced by some
for the patient to receive messages (i.e., tasks and appointment form of opaque algorithmic decision making, we have no control on
requests), set which activities to monitor depending on clinician’s nor clue about its inner functioning. This is very relevant in case
prescription, accept or decline a request sharing certain parts of of recommendations for patients that suffer of chronic illness and
her/his data with a specific clinician, and keep a calendar for tasks that are, usually, elderly. In fact, clinicians need to have control on
and appointments. the feedback given to patients to avoid self-defeating messages that
The clinician makes the prescription of each habit to be mon- may affect patients and/or do not fit with the real needs of a given
itored (i.e., how many steps per day, which and how many pills patient. This motivates the need for moving from black-box to grey-
to take, and which health variable to measure and with which fre- box algorithms, lending themselves to (at least, partial) inspection
quency) through a dedicated web-based application, in which a case and interpretability by human users. In this respect, argumentation
may be defined according to the corresponding clinical pathway, straightforwardly enables algorithms to explain and justify decision
the set of prescriptions to be sent to the SMS, and the clinicians making—both to patients and clinicians.
involved in follow-up of the case. Figure 2 sketches the overall flow This may happen, for instance, by integrating argumentation
of data. The clinician prescribes an activity, the patient receives with NLP techniques to generate explanatory sentences [10]. Ac-
it through the SMS smartphone app, then performs the activity; cordingly, NLP may prove to be invaluable especially in healthcare-
the patient’s wristband monitors the activity, sends data to the related scenarios involving chronic patients and/or elderly people,
smartphone, which are then sent to “the cloud” in which ArgoRec who may be much more accustomed to interact with other people
lives together with the SMS back-end, analysing data and sending (thus, through oral communication) than with technology (that is,
recommendations and feedback. Let us note that how the SMS and through GUI or gestures) [14].
the web application interact is out of the scope of this paper. Argumentation also brings along an interesting opportunity
regarding autonomous learning of recommendation rules, that is, the
4 KEY BENEFITS & CHALLENGES criteria upon which recommendations are provided to the patient.
Experiments with ArgoRec just started with healthy-volunteers In fact, pattern mining techniques are already proficiently employed
in Catalonia. Volunteers were asked to wear a Fitbt charge HR and in many applications of the IoT, where they enable associated rule
to perform their normal activity. In a first period they will be using discovery [21] and user profiling through preferences learning [11].
ArgoRec with the argumentation capability turned off, then it will In this respect, statistical relational learning [12] is a promising
be turned on. Patients’ improvement rate in the two periods will be 1 Clinical studies will start at the end of 2017 in 4 sites: Barcelona, Lleida, Groningen,
measured, as well as efficacy of recommendations—i.e. in terms of and Israel.
HealthRecSys’17, August 2017, Como, Italy J.M. Fernández et al.
source of solutions, since it merges logic with probabilistic models to [2] Punam Bedi and Pooja Vashisth. 2014. Empowering recommender systems using
detect correlations between data despite uncertainty of perceptions, trust and argumentation. Information sciences 279 (2014), 569–586.
[3] Trevor JM Bench-Capon. 2003. Persuasion in practical argument using value-
while exploiting background knowledge to provide explanations based argumentation frameworks. Journal of Logic and Computation 13, 3 (2003),
about the learning process itself—i.e. why and how a given rule has 429–448.
[4] Philippe Besnard, Alejandro Javier García, Anthony Hunter, Sanjay Modgil,
been inferred. Henry Prakken, Guillermo Ricardo Simari, and Francesca Toni. 2014. Introduction
to structured argumentation. Argument & Computation 5, 1 (2014), 1–4.
[5] Lodewijk Bos, Andrew Marsh, Denis Carroll, Sanjeev Gupta, and Mike Rees. 2008.
4.2 Challenges Patient 2.0 Empowerment. In Proceedings of the 2008 International Conference on
Despite argumentation being an active field of research for so long, Semantic Web & Web Services, SWWS 2008, July 14-17, 2008, Las Vegas, Nevada,
USA, Hamid R. Arabnia and Andy Marsh (Eds.). CSREA Press, 164–168.
most of the fundamental results achieved are theoretical. Being [6] Carlos Iván Chesnevar and Ana G Maguitman. 2004. Arguenet: An argument-
interested in applying argumentation in a recommender system to based recommender system for solving web search queries. In Intelligent Systems,
empower complex chronic patients, we move from a theoretical 2004. Proceedings. 2004 2nd International IEEE Conference, Vol. 1. IEEE, IEEE,
282–287.
perspective to the real-world. It is worth noting that the main [7] John Danaher. 2016. The threat of algocracy: Reality, resistance and accommoda-
challenge is moving from a technical perspective (such as finding tion. Philosophy & Technology 29, 3 (2016), 245–268.
[8] Phan Minh Dung. 1995. On the Acceptability of Arguments and its Fundamental
the best logic frameworks) to an organisational and social change in Role in Nonmonotonic Reasoning, Logic Programming and n-Person Games.
case management for both patients and clinicians. In fact, on the one Artif. Intell. 77, 2 (1995), 321–358.
hand, patients have to learn how to interact with suitable devices [9] Paul E Dunne, Anthony Hunter, Peter McBurney, Simon Parsons, and Michael
Wooldridge. 2011. Weighted argument systems: Basic definitions, algorithms,
(i.e., wristband and smartphone or wireless medical devices) and and complexity results. Artificial Intelligence 175, 2 (2011), 457–486.
they have to be confident about the recommendations they receive. [10] Michael Elhadad. 1995. Using argumentation in text generation. Journal of
On the other hand, clinicians have to receive the right information Pragmatics 24, 1–2 (1995), 189 – 220. Argumentation Within Language.
[11] Tom Fawcett. 2015. Mining the quantified self: personal knowledge discovery as
(grey-box approach) to trust the recommendations automatically a challenge for data science. Big Data 3, 4 (2015), 249–266.
generated. What may happen is that, if not correctly motivated, [12] Lise Getoor and Ben Taskar. 2007. Introduction to Statistical Relational Learning
(Adaptive Computation and Machine Learning). The MIT Press.
patients stop to use the self-management system and clinicians [13] Guro Huby, Jenny Holt Brook, Andrew Thompson, and Alison Tierney. 2007.
interrupt prescription of activities through the SMS or checking of Capturing the concealed: Interprofessional practice and older patients’ participa-
the received feedback due to the lack of trust and transparency of tion in decision-making about discharge after acute hospitalization. Journal of
interprofessional care 21, 1 (2007), 55–67.
decision making. [14] Matthias Kranz, Paul Holleis, and Albrecht Schmidt. 2010. Embedded interaction:
Interacting with the internet of things. IEEE internet computing 14, 2 (2010),
46–53.
5 CONCLUSIONS & FUTURE WORK [15] Marco Lippi and Paolo Torroni. 2016. Argumentation Mining: State of the Art
In this paper, we presented the model and architecture of ArgoRec, and Emerging Trends. ACM Trans. Internet Technol. 16, 2, Article 10 (March 2016),
25 pages.
a novel kind of recommender system that relies on argumentation to [16] Kate R Lorig, David S Sobel, Philip L Ritter, Diana Laurent, and Mary Hobbs. 2000.
provide suitable information (rewards, alerts, feedback) to patients Effect of a self-management program on patients with chronic disease. Effective
and clinicians in natural language. ArgoRec has the potential to clinical practice: ECP 4, 6 (2000), 256–262.
[17] S McCann and J Weinman. 1996. Empowering the patient in the consultation: a
sensibly improve patients’ engagement as well as clinicians insights pilot study. Patient education and counseling 27, 3 (1996), 227–234.
into decision making of recommender systems. [18] Richard B Saltman. 1994. Patient choice and patient empowerment in northern
European health systems: a conceptual framework. International journal of health
To substantiate our claim, we just started the evaluation of a services 24, 2 (1994), 201–229.
first proof-of-concept prototype of ArgoRec. The prototype will [19] David Samoocha, David J Bruinvels, Nieke A Elbers, Johannes R Anema, and
be used by healthy volunteers in Catalonia during the summer to Allard J van der Beek. 2010. Effectiveness of web-based interventions on patient
empowerment: a systematic review and meta-analysis. Journal of medical Internet
monitor physical activity (i.e., performed number of steps). Accord- research 12, 2 (2010), e23.
ing to the underlying co-design approach, recommendations will be [20] Nancy Tomes. 2007. Patient empowerment and the dilemmas of late-modern
analyzed by the users as well as by clinicians from Hospital Santa medicalisation. The Lancet 369, 9562 (2007), 698–700.
[21] Chun-Wei Tsai, Chin-Feng Lai, Ming-Chao Chiang, Laurence T Yang, et al. 2014.
Maria in Lleida (Catalonia, Spain). The corresponding feedback will Data mining for Internet of Things: A survey. IEEE Communications Surveys and
be used to improve the system and get it ready to be used in the Tutorials 16, 1 (2014), 77–97.
[22] Frans H Van Eemeren, Rob Grootendorst, Ralph H Johnson, Christian Plantin,
CONNECARE project with patients from Barcelona, Lleida, Gronin- and Charles A Willard. 2013. Fundamentals of argumentation theory: A handbook
gen and Israel. In particular, two case studies will be considered: of historical backgrounds and contemporary developments. Routledge.
(1) Community-based management of complex chronic patients, [23] Eloisa Vargiu, Juan Manuel Fernández, Felip Miralles, Isaac Cano, Elena Gimeno-
Santos, Carme Hernandez, Gerard Torres, Jordi Colomina, Jordi de Batlle, Rachelle
and (2) Preventive patient-centered intevention in complex chronic Kaye, Bella Azaria, Shauli Nakar, Maarten M.H. Lahr, Esther Metting, Maarten
patients undergoing elective major surgical procedures. Jager, Hille Meetsma, Stefano Mariani, Marco Mamei, Franco Zambonelli, Felix
Michel, Florian Matthes, Joanna Goulden, John Eaglesham, and Charles Lowe.
2017. Integrated Care for Complex Chronic Patients. In ICIC 17 - 17th International
ACKNOWLEDGMENTS Conference on Integrated Care "Building a platform for integrated care: deliver-
ing change that matters to people". May 8-10, 2017, Dublin (Ireland). Internation
The work is supported by the CONNECARE (Personalised Con- Foundation for Integrated Care, Dublin, Ireland.
nected Care for Complex Chronic Patients) project (EU H2020-RIA,
Contract No. 689802).
REFERENCES
[1] Reinhard Angelmar and Philip C Berman. 2007. Patient empowerment and effi-
cient health outcomes. Financing sustainable healthcare in Europe: new approaches
for new outcomes 1, 2 (2007), 3.