Personalization of Computer-Based Technologies for Autism: An Open Challenge for Software Engineering?

Autism Spectrum Disorder (ASD) is neurodevelopmental condition characterized by social interaction and communication difficulties, along with narrow and repetitive interests. Being an spectrum disorder, ASD affects individuals with a large range of combinations of challenges along dimensions such intelligence, social skills, or sensory processing. Hence, any computer-based technology for ASD ought to be personalized to meet the particular profile and needs of each person that uses it. Within the realm of Software Engineering, there is an extensive body of research and practice on software customization whose ultimate goal is meeting the diverse needs of software stakeholders in an efficient and effective manner. These two facts beg the question: Can computer-based technologies for autism benefit from this vast expertise in software customization? As a first step towards answering this question, we performed an exploratory study to evaluate current support for customization in this type of technologies. Our study revealed that, even though its critical importance, customization has not been addressed. We argue that this area is ripe for research and application of software customization approaches such as Software Product Lines.

READ FULL TEXT VIEW PDF
POST COMMENT

Comments

There are no comments yet.

Authors

page 1

page 2

page 3

page 4

05/08/2018

A Focused Mapping Study on Customization in Interactive Technologies for Autism

Autism Spectrum Disorder (ASD) is neurodevelopmental condition character...
12/14/2017

Fourteen Years of Software Engineering at ETH Zurich

A Chair of Software Engineering existed at ETH Zurich, the Swiss Federal...
06/11/2014

A Hitchhiker's Guide to Search-Based Software Engineering for Software Product Lines

Search Based Software Engineering (SBSE) is an emerging discipline that ...
05/09/2019

A Road to Bio-inspired Software Engineering

Software production research is quickly evolving on two parallel approac...
01/05/2020

Unleashing the Potentials of Immersive Augmented Reality for Software Engineering

In immersive augmented reality (IAR), users can wear a head-mounted disp...
10/09/2020

Research, Develop, Deploy: Building a Full Spectrum Software Engineering and Research Department

At Sandia National Laboratories, the Software Engineering and Research D...
02/23/2017

Building Usage Profiles Using Deep Neural Nets

To improve software quality, one needs to build test scenarios resemblin...
This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

1 Introduction

Autism Spectrum Disorder (ASD) is neurodevelopmental condition characterized by social interaction and communication difficulties, along with narrow and repetitive interests. ASD affects individuals in multiple and combined ways along areas such intelligence, social skills (e.g. unable to interpret non-verbal cues), or sensorial processing (e.g. sensitivity to noise or lights) [ANYAS/ElKalioubyPB06]. In the autism community, a common saying is: ”if you’ve met one person with autism you’ve met one person with autism”111Quote authored by Stephen Shore, http://www.autismasperger.net/.. This entails that individuals with autism have unique sets of challenges and needs that must be addressed to help their development and integration to society.

There is an extensive and long standing research on using computer-based systems, that spreads over more than four decades [Kientz2013], whose driving goal is to support the needs of people with autism and their families. Currently, digital libraries have hundreds of articles on the subject. This research has been summarized to certain degree in many literature reviews and surveys studies (e.g. [SAGE/GrynszpanWPG14, Stephenson2015]). However, and despite its critical importance, there has not been a study on how this type of computer-based systems handle customization, also referred to in the autism community and literature as personalization . In this paper, we address this issue by analyzing the customization capabilities of approaches published over the last five years. Furthermore, we want to collect information about the strength of the empirical evidence that supports each of the approaches. In other words, the types of formal research studies that have been performed with them.

Our exploratory study indeed corroborated the lack of research on customization for computer-based technologies for autism and the slim empirical evidence that supports them. Based on these findings, we argue that this area is ripe for research and application of advanced customization approaches such as Software Product Lines. We conclude by sketching some first challenges in this area.

S2-PUCom/SitdhisanguanCDO12 S3-PUCom/Keay-BrightW12 S9-DISC/HailpernHKBK12 S13-RosenbloomMWM16 S14-MechlingAFB15 S15-LiuSVS17 S16-TorradoGM17 S17-CabiellesPPF17 S18-ChevalierMIBT17 S19-YingSA16 S20-KhanPHHSM16 S21-SharmaSAVHKTR16 S22-MagriniSCC16 S23-HulusicP16 S24-SturmPP16 S25-ChevalierIMT16 S26-LandowskaS16 S27-SkillenDNB16 S28-KhoslaNC15 S29-MeiMQ15 S30-VoonBMJA15 S31-QidwaiSHM14 S32-GruarinWB13 S33-DimitrovaVB12

2 Autism Background

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by impaired social interaction and communication, and restricted and repetitive behavior [APA]. ASD is diagnosed in at least 1% of the population, and diagnoses are more common amongst males than females [Baird2006]

. Autism can have profound impact upon learning and it is estimated that 54% of individuals with autism also have intellectual disability/learning disability (Center for Disease Control CDC

222Center for Disease Control (CDC) https://www.cdc.gov/ncbddd/autism/data.html).

Today, no medical treatment is available for the core symptoms of autism. Early intervention programs, usually aimed at children from 0 to 6 years old, have been demonstrated effective for supporting the development of a relevant percentage of children with autism. The most effective programs have a behavioural base (e.g. Applied Behaviour Analysis (ABA) [Cohen2006]) or cognitive-behavioural base (e.g. Early Start Denver Model (ESDM) [Rogers2010]). Within this later program, for example, there is a developmental curriculum (the ESDM checklist) that is highly personalised for each child with autism and intervention objectives are redefined every three months in order to adapt to the child progress.

Intervention programs in autism can be classified as focused-intervention (FI) programs 

[Wong2015], usually designed for improving a particular ability (or a reduced set of abilities) or comprehensive treatment models (CTM) that are much wider and are based on a holistic approach of the child development [Odom2010]. Some of these programs use technologies as a basis for documenting the child progress, and some other use technology for very particular tasks. However, none of these programs are genuinely based on any particular technology. When available, innovative technologies are used for Focused Interventions rather than as Comprehensive Treatment Models. Most research evidence available on technologies for ASD rely mainly on the use of particular communicator apps on tablet devices while the evidence on other areas seems to be anecdotal or at least not enough explored [Lorah2014].

3 Exploratory Study

The goal of our exploratory study is to gauge at the support of customization on computer-based technologies for autism. Hence, to retrieve the relevant literature we performed a search using Web of Science333http://apps.webofknowledge.com/. We employed this search engine because its advanced query capabilities and because it indexes all the publication outlets on autism and technology. In this search, in addition to the term customization, we also use the term personalization as it is also commonly used in the autism literature. We constrained our query to sources published in the last five years. The query we employed was444TS=Topic Search, ASC=Autism Spectrum Condition:

(TS=((Autis* OR ASD OR ASC OR "Asperger Syndrome" OR "Pervasive Developmental Disorder" OR PDD*) AND (Technolog* OR Computer* OR Virtual* OR Robot*) AND (Custom* or Personali*))) AND (Search Language = English) AND (Document type= Article) AND (Timespan= 2012-2017) AND (Indexes: SCI-EXPANDED, SSCI, A&HCI, CPCI-S, CPCI-SSH, ESCI, CCR-EXPANDED, IC.)

Inclusion and exclusion. The basic criterion for inclusion in our study was a clear application of a computer-based technology for supporting a therapy or intervention in relation to autism, where individuals with autism participated in the design, validation or evaluation of the technology. The criteria to exclude papers in our study was: i) papers which did not describe a technology that supports any intervention or therapy (e.g. paper that describes biosignal monitoring tool), ii) individuals with autism were not involved at any stage of design, validation or evaluation of the technology, iii) papers not written in English, iv) vision or position papers that had no implementation to back them up, v) graduate or undergraduate dissertations and thesis, and vi) non peer-reviewed documents such as technical reports.

During the screening process we looked for the search terms in the title, abstract and keywords and whenever necessary at the introduction or at other places of the paper. The decision on whether or not to include a paper was most of the times straightforward, in other words, a clear application of computer-based technologies to autism with the participation of individuals with the condition was easily drawn.

The search query obtained 179 sources. After a careful sieving using the inclusion and exclusion criteria, our search produced 24 primary sources that we analyzed in further detail as presented in Table 1, which shows the type of technology used, the forms of interactions, and the support for customization. From this table, we can observe : i) the pre-eminent technologies are mobile and smartphone devices, followed by robots; ii) the predominant use of touch screen as form of interaction; and iii) the dire lack of customization support, with 11 out of the 24 actually providing examples of customization and not simply mentioning it as desirable property.

Primary source Technology Interaction forms Customization support
S2-PUCom/SitdhisanguanCDO12 tangible user interfaces object manipulation no support provided
S3-PUCom/Keay-BrightW12 multi-touch screen touch screen no support provided
S9-DISC/HailpernHKBK12 speech recognition and visual feedback monitors and microphones no support provided
S13-RosenbloomMWM16 smartphone screen interaction customization for prompts, recording, data monitoring
S14-MechlingAFB15 video recording and playing video watching highlights importance of custom-made videos
S15-LiuSVS17 smart glasses, augmented reality, games eye gaze, movement sensor data, game difficulty and rewards
S16-TorradoGM17 smartwatches, smartphones app, smartwatch screen authoring tool for coping strategies
S17-CabiellesPPF17 tablets tablet screen sequences, words
S18-ChevalierMIBT17 avatars, robots tacticle computer game no support provided
S19-YingSA16 mobile devices, smartphone touch screen, mobile phone default avatars with personal pictures
S20-KhanPHHSM16 mobile devices touch screen, mobile phone no support provided
S21-SharmaSAVHKTR16 gesture detection gestures no support provided
S22-MagriniSCC16 gesture detection, body tracking gestures, movement sound selection for gestures and movement
S23-HulusicP16 web game web page not support provided
S24-SturmPP16 mobile devices touch screen, mouse therapy session contents
S25-ChevalierIMT16 robots gestures, movement 3 groups based on kinetic and propioception profiles
S26-LandowskaS16 tablets touch screen activities plan
S27-SkillenDNB16 smartphone touch, visual colors scheme, logging info, geo-fences
S28-KhoslaNC15 robots speech, touch screen activities and care services
S29-MeiMQ15 virtual reality, body tracking motdio detection, touch screen, screen avatars in game
S30-VoonBMJA15 mobile devices, tablet touch screen, sound, visual images, stepwise description of tasks
S31-QidwaiSHM14 robots motion detection, tablet play activities
S32-GruarinWB13 sensors tangible carpet tasks and photographs
S33-DimitrovaVB12 robots movement no support provided
Table 1: Technology, interaction forms, and customization support summary

Empirical evidence support. This refers to the empirical evidence that supports each of the sources identified, which we describe in terms of the research designs (or experimental designs) employed. We have considered the two principal types of design that are applied in most technological studies [Kientz2013, WohlinExperimentalSE12]. Single subject research design refers to research in which the subject serves as his/her own control, rather than using another individual/group. Group research design refers to research where one group of participants (treatment group) is compared to another group (control group) with participants in both groups balanced around variables such as age, IQ or severity of autism symptoms around social communication or restrictive/repetitive behaviours. Additionally, we classify also the availability of the resources for replication, for instance open sources and documentation. For this latter category, we use none, partial, and full depending on the degree of availability.

Table 2 summarizes our findings. It is immediately clear that the majority of sources utilize the most basic type of design, single subject, with a very reduced number of participants, and only in one case all the information for replication is provided. Furthermore, five sources utilise group designs but the lack of available details hinder their replication. Finally, other five sources did not make any empirical evaluation or at least they did not describe the study in enough detail.

Single subject design No Level
S27-SkillenDNB16 16 None
S3-PUCom/Keay-BrightW12 13 Partial
S17-CabiellesPPF17 11 Partial
S29-MeiMQ15 10 Partial
S19-YingSA16 5 Partial
S14-MechlingAFB15, S22-MagriniSCC16 4 Partial
S26-LandowskaS16 2 Full
S9-DISC/HailpernHKBK12, S15-LiuSVS17, S16-TorradoGM17, S17-CabiellesPPF17, S20-KhanPHHSM16 2 Partial
S28-KhoslaNC15 2 None
S32-GruarinWB13 1 Partial
S13-RosenbloomMWM16 1 None
Group design No Level
S2-PUCom/SitdhisanguanCDO12 32 None
S33-DimitrovaVB12 22 None
S18-ChevalierMIBT17, S25-ChevalierIMT16 13 Partial
S21-SharmaSAVHKTR16 10 Partial
Not possible to determine No Level
S1-IWVR/ParesCDFFGS04 11 Partial
S31-QidwaiSHM14 NA Partial
S23-HulusicP16, S24-SturmPP16, S30-VoonBMJA15 NA None
Table 2: Empirical support summary

4 Challenges for Software Engineering

One of the leading approaches for software customization are Software Product Lines (SPLs) which are families of related systems whose members are distinguished by the set of features they provide, where a feature is an increment in functionality [DBLP:journals/tse/BatorySR04, SPLE]. A key concept in SPLs is variability which is the capacity of software artifacts to vary. Several forms of variability models have been proposed that succinctly and formally express all the desired combination of features that the products of an SPL may have. Our study has revealed the following two open challenges:

  • Develop user profile models that formally describe all the variations that persons with autism may have along dimensions like sensory needs, intellectual disability, etc. taking information from the standard battery of tests used to diagnose the condition.

  • Provide tool support to collect and analyze data that integrates with the standard workflow developers use (e.g. a plug-in for Android studio when developing apps) for improving the empirical evaluation of this type of technologies.

5 Acknowledgments

This study has been posible thanks to the funding received under the grant agreement 2015-1-ES01-KA201-015946 of the Erasmus+ Program of the European Union and the project Recherche interdisciplinaire sur les systèmes logiciels variables sponsored by the École de technologie supérieure, University of Québec, Canada.

References