Behaviour & Information Technology

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/tbit20

User-specific touch interfaces: a viable solution for

an aging society?

Avner Shahal, Robert P. Spang, Michael Minge, Carola Trahms & Jan-Niklas
Voigt-Antons

To cite this article: Avner Shahal, Robert P. Spang, Michael Minge, Carola Trahms & Jan-
Niklas Voigt-Antons (2021): User-specific touch interfaces: a viable solution for an aging society?,
Behaviour & Information Technology, DOI: 10.1080/0144929X.2021.1906322

To link to this article: https://doi.org/10.1080/0144929X.2021.1906322

View supplementary material

Published online: 01 Apr 2021.

Submit your article to this journal

Article views: 45

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tbit20
BEHAVIOUR & INFORMATION TECHNOLOGY
https://doi.org/10.1080/0144929X.2021.1906322

User-specific touch interfaces: a viable solution for an aging society?

a,c b a b,e b,d
Avner Shahal , Robert P. Spang , Michael Minge , Carola Trahms and Jan-Niklas Voigt-Antons
a
Chair of Cognitive Psychology and Cognitive Ergonomics, TU Berlin, Germany; bQuality and Usability Lab, TU Berlin, Germany; cmedneo
GmbH, Berlin, Germany; dGerman Research Center for Artificial Intelligence (DFKI), Berlin, Germany; eArchaeoinformatics – Data Science, Uni
Kiel, Germany

ABSTRACT ARTICLE HISTORY

Touch interaction has established a dominating role in the realisation of Human–Machine Received 24 July 2020
Interfaces. However, to be able to use touch effectively and efficiently, users have to comply Accepted 16 March 2021
with particular prerequisites. Due to age-related changes, such as the decline of tactile accuracy
KEYWORDS
and speed, especially elderly users often struggle with the touch modality. Interfaces that adapt Adaptive systems; elderly
to specific user characteristics could be a promising solution to overcome this problem. users; touch interaction;
Notwithstanding the advantages of adaptive systems, perceived changes in the user interface design for all
can reduce the system’s predictability and transparency. The present study compares three
approaches concerning the adaptation of touch button sizes: no adaptation and adaptation
with visible and invisible feedback. Results show that especially elderly users substantially
benefit from an adaptive approach. Furthermore, data shows that the type of adaptation
supports different usage goals. While adaption with visual feedback enables a higher interaction
speed, invisible adaptation leads to a higher degree of accuracy.

1. Introduction not surprising that some researchers regard touch as

‘[…] the new click’ (Saffer 2008). The benefits and
1.1. ICT and touch
increasing dissemination of touch notwithstanding, it
Information and Communication Technology (ICT) is does not come without its costs (e.g. Hertzum and Horn-
changing the way we interact with the world. Many bæk (2010)). In order to be able to use touch interfaces
basic aspects of everyday life have become essentially effectively and efficiently, users have to comply with cer-
intertwined with the use of ICT, to the point in which tain prerequisites, such as eyesight, manual dexterity,
the absence of the internet, personal computers, or hand- and cognitive processing ability. Not being able to fulfil
held devices could barely be imagined. Within the ever- such requirements may lead to the exclusion of certain
growing dependence on ICT and because of their intui- user groups from touch devices.
tiveness, touch displays have been regarded as the cur-
rent dominating implementation of Natural User
1.2. Aging
Interfaces (NUIs; Wigdor and Wixon (2011)). Touch
screens combine a visual display with input option and, As adults grow older, a general decline in cognitive, sen-
thereby, allow for direct manipulation of information. sory-motor and perceptual abilities occurs. Older adults’
Touch interaction is currently the interaction-modality information processing ability generally becomes slower
of choice for most of the current Human-Machine-Inter- and their reaction time consequently increase (Ketcham
faces (HMI) (Walker, Philbin, and Fisk 1997; Watson and Stelmach 2004; Laugwitz, Held, and Schrepp 2008;
et al. 2013) and has become a state-of-the-art of HMI Craik and Salthouse 2011). The execution of movement
(Saffer 2008; Claypoole, Schroeder, and Mishler 2016). generally becomes slower and more variable (Ketcham
Although touch displays do not necessarily promise a and Stelmach 2004; Laugwitz, Held, and Schrepp 2008).
more effective or efficient interaction (e.g. Goldstein This decline in motor control is accompanied by a decline
et al. (1999); Hertzum and Hornbæk (2010); Gunawar- of perceptual abilities, such as eyesight and tactile percep-
dana, Paek, and Meek (2010)), research has shown, that tion (Walker, Philbin, and Fisk 1997; Ziefle 2001; Wright
touch displays can offer higher hedonic ratings than tra- et al. 2000; Lavie and Meyer 2010; Dunn et al. 2015).
ditional interfaces (Mathiowetz et al. 1985; Ziefle 2001; Research has shown that elderly people understand
Leonardi et al. 2010; Watson et al. 2013). It is therefore many touch gestures and come to terms with the use of

CONTACT Robert P. Spang [email protected]

© 2021 Informa UK Limited, trading as Taylor & Francis Group
2 A. SHAHAL ET AL.

touch interfaces (e.g. Stößel, Wandke, and Blessing et al. 2015). Although adaptive systems can be regarded
(2010)). A large body of research has shown, however, as competent extensions of conventional HMIs, their
that older users often struggle with touch interaction implementation is scarce. Beyond the substantial
(Fitts 1954; Wright et al. 2000; Chung et al. 2010; Barnard increase in development-complexity and costs involved
et al. 2013). In particular, problems occur in case of pre- in the implementation of adaptive systems, it has been
cision-relevant touch functions, such as time-critical ges- suggested that adaptive systems risk violating basic
tures (e.g. double tap, long press) and small icons (Smith, usability principles by reducing the system’s predictabil-
Sharit, and Czaja 1999; Ziefle and Bay 2005; Darroch et al. ity and transparency and thus hindering users’ percep-
2005; Naumann, Wechsung, and Hurtienne 2010; Choi tion of control over the system (Gena 2005).
and DiNitto 2013; Claypoole, Schroeder, and Mishler
2016). In conclusion, a substantial body of research con-
1.4. Adaptive systems and age
sistently shows a general decline in sensory-motor, per-
ceptual, and cognitive abilities with age. This Lavie and Meyer (2010) investigated different levels of
accumulated evidence is highly relevant when consider- adaptivity and task difficulty in familiar and unfamiliar
ing the interaction-characteristics of elderly users with situations, comparing younger users (M = 25.7 years)
technological appliances such as computers and when with elderly persons (M = 58.6 years). They found that
designing user interfaces that should be accessible to the older age group benefited more from an adaptive
elderly users (Kobayashi et al. 2011). interface. On the one hand, it relieved load caused by
the task itself, but at the same time, participants were
more handicapped by such a system in non-routine
1.3. Adaptive systems
situations, where the costs of overriding the adaptive
Although the differences between users have been suggestion were much higher. Thus, one of the main
acknowledged, development has all too often focused drawbacks of adaptive interfaces was the increasing
on the ‘average’ user (young adults) (Bühler and Stepha- costs through manual interventions needed to override
nidis 2004; Arnett 2008; Hanel and Vione 2016). To suggestions of the adaptive system. This aspect of adap-
design touch interfaces accessible for older users and tive systems is especially problematic because an over-
users with motor impairments, developers and commu- ride requires the user to have a good mental concept
nities have set their goal to use the ‘Design-for-All’ and understanding of the system (to allow for predict-
(DfA) strategy. It aims at applying principles, methods, ability, transparency, and error handling). As the adap-
and tools in the development of interactive systems to tive aspects of a system assist users in achieving their
design interfaces that are suitable for most potential interaction goals, such systems implicitly hinder the
users without modification or individually adaptable refinement of a precise mental concept required for
to different users (Stephanidis 2000; Bühler and Stepha- manual intervention. In their study, Lavie and Meyer
nidis 2004). Adaptive touch interfaces, which can found a larger performance drop for users accustomed
change certain interaction aspects according to the to a highly adaptive system than users in a mixed-
users’ characteristics, could provide a promising sol- initiative interaction condition, assuming skill degra-
ution to these issues in the future. By acquiring data dation over time as the cause of the difference. The
concerning the individual’s interaction properties, authors suggested keeping adaptivity at an intermediate
adaptive systems can adjust HMI properties, such as level, thus keeping the human in the loop and allowing
the size of interactive elements. Accommodating the the user to establish a mental concept of the system.
needs of users who would otherwise fail to meet the A possible solution for this problem could be an adap-
accuracy required to interact with regular touch displays tation that only happens behind the scenes and that is not
efficiently would provide each user with the most suit- visible for the user in the graphical user interface. For
able interface, allowing a plurality of solutions for a instance, the apparent size of a button could be kept con-
large and heterogenic user group. By adapting to the stant during adaptation, whereas only the touch-sensitive
user’s characteristics and abilities, adaptive systems interaction area assigned to this button, will be adapted.
offer a solution that allows achieving DfA goals. Adap-
tive language models have, for example, shown to be
1.5. Research questions
helpful when implementing a virtual keyboard for text
entry on touch displays (Goodman et al. 2002). Simi- From the review of the current literature, the effects of
larly, input speed and accuracy are a promising way to both subjective and objective aspects of the visibility
significantly improve through the adaptation of a virtual of adaptation have yet to be explored. They are crucial
keyboard to the user’s interaction characteristics (Mock to the proper and efficient implementation of adaptive
 BEHAVIOUR & INFORMATION TECHNOLOGY 3

solutions in a plurality of scenarios and systems. From counterbalanced, 2 × 3 (age group, adaptation type)
the above, the following questions arise: (I) Would mixed design with repeated measures. Subjective
users prefer adaptive interfaces in which the change measures followed every exposure to the stimulus and
due to adaptation is visible over interfaces with similar were analysed in the same fashion. The design com-
adaptive characteristics, in which changes are not vis- prised the between-subjects factors age group (young
ible? (II) Would performance involving visible adap- 20–30, elder 64–84 years) and adaptation type (I, II, III).
tation be objectively better than with the invisible
adaptation paradigm? (III) Would elderly users’ prefer-
2.1. Participants
ences and performances differ or be the same as those of
younger adults? Participants were recruited using an online recruiting
tool at Technische Universität Berlin and via the Senior
Research Group’s meetings in Berlin, Germany. The lat-
1.6. Aims of current research
ter is an informal interest group of mostly retired
In the following study, we explore the effects of the visi- researchers. Because we got in touch via email, we
bility of adaptation in the realm of touch devices, while assume not to have contacted and completely non-tech-
focusing on the often neglected elderly user group and nology-savvy persons. While the online recruiting tool
comparing it to young adults, for which most touch is free to use for everybody, most seniors were recruited
interfaces are being conceived. We aim to explore the through the Senior Research Group’s mailing list. Cri-
benefit of adaptive touch displays concerning the elderly teria for participation were normal or corrected to nor-
user group and to gain insights on the effects of the visi- mal vision, right-hand dominance and an age of 18 to 30
bility of adaptation on objective and subjective aspects for the young-age group, or above 65 for the elderly
of adaptive systems. group. Participants were asked to register if they
match these mandatory criteria only.
A total of 48 participants, 24 in each group, took part
1.7. Hypotheses
in the experiment. The younger-age sample ranged in
This work’s goals are twofold. First, the study will explore age from 20 to 30 years (M = 24.8, SD = 2.65), the
the impact of adaptive user interfaces on elderly users, older-age group ranged in age from 64 to 84 (M =
using objective and subjective measures to evaluate 71.4, SD = 5.85). 54% of the participants identified
age-group differences regarding a system’s adaptivity. themselves as female; no one identified as non-binary.
We hypothesise that elderly users show less manual dex- Participants were able not to answer, not to respond,
terity than younger users as well as less technical affinity. and to withdraw from the study at any time. Each par-
Taking the user’s mental representation into con- ticipant received either partial course credit or an incen-
sideration, the second goal of this study is to explore tive of ten euro as compensation for participation.
the characteristics of different adaptation options: vis-
ible and invisible. Here we hypothesised, that objective
2.2. Apparatus
and subjective aspects of users’ interaction will be better
in the adaptation conditions than the no-adaptation Touch-interaction tasks were performed on a 9.7-inch
condition. Furthermore, that invisible adaptation is Apple iPad Air 2. The iPad was placed on a table in front
more advantageous than visible adaptation when com- of the participants at an angle of 25 degrees from the table’s
paring adaptation methods. As a possible interaction surface, using a commercially available Ikea iPad holder.
of age group and adaptation type is not supported by lit- Markings on the table were used to hold the iPad position
erature and cannot be intuitively suggested, hypotheses constant across all participants. The iPad screen was
focus on the previously-stated main effects. cleaned between participants to avoid an accumulation
of finger marks on the display. The experiment was con-
ducted at Technische Universität Berlin. Lighting, noise
2. Methods
levels, and room temperature were controlled in order to
We evaluate the characteristics of different adaptation keep conditions similar to each participant.
types and compare them with the control condition
(no-adaptation) in a laboratory setting. Both young
2.3. Stimuli
and elderly users performed an interaction task present-
ing simple touch-related exercises on a tablet. Objective Interaction Game To test participants’ interaction charac-
measures of users’ interaction with the system were col- teristics with different adaptation types, a spelling game
lected and analysed using a fully-crossed, was used as a primary task, in which participants spell
4 A. SHAHAL ET AL.

different words at a time. The system recorded multiple multiple interactions. Following these requirements, sev-
aspects of participants’ touch-characteristics and their eral different algorithms were developed and tested via
interaction with the task. The interaction game was pilot studies. The resulting algorithm allowed changing
based on ‘PflegeTab’, a programme developed by the the size of interaction elements according to the user’s
‘Quality and Usability Lab’ at Technische Universität Ber- accuracy in the past five interactions. This window size
lin (Antons et al. 2016; Nordheim et al. 2017). Intended for resulted from several test iterations, allowing for continu-
cognitive rehabilitation and emotional activation of ous changes while not updating the hit-box too rapidly or
dementia patients, PflegeTab can be used on the Apple too sluggishly. After testing several bidirectional adap-
iPad tablet. It consists of multiple in-app games, such as tation algorithms, an adaptation was chosen, which only
picture matching, quiz, and spelling. This study concen- allowed increasing the button size. The initial size of the
trated solely on the spelling game of the PflegeTab. The interaction elements was 25 points (of width and length;
latter enables users to spell predefined words, using their this translates to 50px2 or 0.48 cm2 ). The maximum size
fingers to select, by tapping, the required letters. Letters was constrained to 150 points (300px2 / 2.89 cm2 ). In
could only be selected in the correct order so that partici- both adaptation conditions (Visible and Non-Visible),
pants did not have to plan multiple spelling strategies. the same algorithm for the button size specification was
Upon the initialisation of each task, all letters appear sim- implemented. The update rule for the button sizes of the
ultaneously, together with the target word, which should final algorithm is described in Equation (1). It defines
be spelled. Each letter is assigned a random position on the current button size Ct as the sum of the previous but-
a predefined screen area, avoiding overlap with the pos- ton size Ct−1 and the arithmetic product of the predefined
ition of a different letter. Both the target word and the step size Z and the miss-ratio λ. The latter resulted from
letters that have already been selected were always visible, the ratio of correct to incorrect selections in a sliding win-
thus reducing the requirement of memorising the word or dow of five interactions. As described above, the pre-tests
spelled letters. Using this procedure, we minimised poten- showed a sliding window of the past five interactions to
tial insecurities for the participants: instead of spelling result in the most satisfying adaptation. Here, the number
words, the challenge was primarily a visual search task. of incorrect selections out of the past five trials is used for
Furthermore, all participants hold a German school the computation of λ. Likewise, the step size Z = 2 pro-
degree so general language comprehension can be vided sufficient change in button size while maintaining
assumed. To further minimise potential insecurities and enough room for change in upcoming iterations.
the feeling of being observed, the experimenter sat down
ct = ct−1 + lz (1)
with the participants at a 90 degrees angle, unable to
sight the iPad screen. After selecting the last letter of the
current word, a motivation screen was shown for 5
2.4. Measurement tools
seconds, after which the next task–a new word–appeared.
Words were randomly picked out of a database consisting Subjective experienced effort questionnaire The ‘Subjek-
of 290 words, with a length of 2 to 12 letters. tive erlebte Anstrengung’ (English: Subjective Experi-
Adaptation Types In order to allow testing different enced Effort; Eilers, Nachreiner, and Hänecke (1986))
adaptation strategies, we focused on the spelling game was administered immediately after completing each
in the PflegeTab environment and adapted its character- condition. This measurement consists of a vertical
istics. The following adaptation types were implemented scale 10-cm long, with twelve labels of units depicted
in the game: No-adaptation of button size (control). Vis- on the scale. Labels bared the numbering 0 (no effort
ible adaptation, in which the button size visibly changed required) to 220 (extreme effort required) with non-lin-
when adaptation took place. Invisible adaptation, in ear steps of 20. Short definitions for certain effort levels
which the visible button stayed constant, while the inter- were available in German in the vicinity of the scale.
action area changed stepwise according to previous Usability experience questionnaire Participants were
interactions with the user interface during the same con- asked to fill out the German version of the ‘User Experi-
dition. Figure 1 compares the three experimental con- ence Questionnaire’ (Laugwitz, Held, and Schrepp 2008)
ditions before (t0) and after (t1) several interactions. after each condition. The UEQ consists of 26 bipolar
Adaptation mechanism The adaptation mechanism items (semantic differential), divided into six subscales:
needed to fulfil several requirements. Most importantly, Attractiveness, Perspicuity, Efficiency, Dependability,
it needed to facilitate a change in the target letters’ button Stimulation, and Novelty.
size according to the user’s accuracy in the past inter- Box and block test of manual dexterity To test partici-
actions. Furthermore, it needed to support an iterative, pants’ manual dexterity, the box and block test (Mathio-
stepwise adaptation to accumulate small changes over wetz et al. 1985) was administered at the beginning of
 BEHAVIOUR & INFORMATION TECHNOLOGY 5

Figure 1. Adaptation both at the beginning of the interaction with the system, before adaptation took place (t0), as well as after
iterative spelling tasks (t1) for each of the experimental conditions. The dashed border around the letters depicts the non-visible inter-
action area.

the experiment. As no German version of the test is dexterity (Mock et al. 2015). After completion, participants
available, instructions were translated into German were asked to fill in the technical affinity questionnaire.
and reviewed by three native German speakers. Subsequently, participants received the spelling task
Technical affinity questionnaire The technical affinity instructions and were allowed to get acquainted with it
questionnaire (Karrer et al. 2009) was administered in through a practice phase, which lasted three words. The
the German language to acquire participants’ attitudes practice phase was performed with the no-adaptation con-
towards technical appliances. This questionnaire includes dition in order to avoid any pre-experience with the adap-
19 phrases, to which participants were asked to give their tation. Upon completion, three runs were performed, in
opinion on a Likert scale of 1 (completely disagree) to 5 which participants were requested to interact with the spel-
(completely agree). The questionnaire’s items are divided ling task. Condition-order was counterbalanced between
post-hoc into four predefined-subscales: Enthusiasm, participants. Participants were not informed about the
Competence, Negative Beliefs, and Positive Beliefs. existence of different conditions and were unaware of the
Touch information A plurality of touch-parameters condition type they were interacting with. The game was
was saved during participants’ interaction with the spel- pre-set to allow a 5-minute interaction per condition and
ling game. These parameters included general infor- stopped automatically. After selecting the last letter of the
mation regarding the task which the participant was current word, a motivation screen (e.g. ‘well done!’, but in
performing (e.g. word length, adaptation type, button German) was shown for 5 seconds, after which the next
x and y dimensions, button coordinates on the screen). word appeared. After each condition, participants were
Additionally, they included information regarding par- requested to fill out the SEA and the UEQ. Upon complet-
ticipant’s interaction with the system (e.g. hit or miss of ing the questionnaires, participants were allowed to take a
the target button, the timestamp for each interaction). short break before beginning the next condition. After per-
Furthermore, the system continuously logged general forming all three conditions, a short, non-standardised
data about the specific task (e.g. word length, button post-session interview was held, and participants received
dimensions, and coordinates). compensation for their participation (see Figure 2).

2.5. Procedure 2.6. Operationalisation

Upon arrival and after giving their informed consent, par- To gain insight into the use of adaptation, the last button
ticipants performed the Box and Block Test of manual size was calculated for each participant for each
6 A. SHAHAL ET AL.

81.6 − 0.30 ∗ age, with age in years. The simple linear

regression provides a sense as to how the manual dex-
terity decreases with increasing age. Most importantly,
results show that participants of the different age groups
in this experiment did differ in their manual dexterity.
Technical affinity The technical affinity questionnaire
(Karrer et al. 2009) was tested regarding each of the four
previously-mentioned sub-scales. No significant differ-
ence between age groups was found regarding the
Enthusiasm, Competence, and Positive Beliefs sub-
scales. A significant difference was found between the
scores of the old age group (M = 3.67, SD = 0.75) and
the young age group (M = 3.24, SD = 0.54);
t(41.8) = 2.27, p = .029, d = 0.65 regarding the Negative
Beliefs subscale. This difference suggests that, although
age groups did not differ regarding the first three sub-
Figure 2. Flowchart of the study protocol. scales, older participants expressed more negative
beliefs regarding the use of technical appliances than
adaptation condition. The No-Adaptation condition was participants of the younger group.
of no relevance for this calculation, as the button size did
not change during this condition. The ratio between the
3.2. Statistical design and method
total number of correct selections (‘hit’) and the total
number of touches (‘total’) was used as a measurement Statistical Analyses For the range of adaptation, as
of accuracy. This ratio was computed for all participants measured using the button-end-size, a mixed-design
and each condition. Finally, we surveyed the spelling analysis of variance with age group (young, old) as the
speed as a measure of interaction efficiency. Spelling between-subjects factor and the two experimental con-
speed was defined as the mean of correct letters tapped ditions (Visible Adaptation and Non-Visible Adap-
per second and can be pictured as ‘fluidity of typing’. tation) as the within-subjects factor was conducted. As
As to the subjective measurements, the workload was the button-size of the no-adaptation condition remained
measured according to the previously-mentioned SEA 25 points, this condition was not included in the analysis
scale, and a usability measurement was performed of variance mentioned above. For all other dependent
using the UEQ questionnaire. variables (i.e. interaction accuracy, interaction speed,
the subjectively experienced effort questionnaire and
the six UEQ subscales), a multivariate analysis of var-
3. Results iance was performed first, followed by independent
mixed-design analyses of variance for each dependent
All participants chose to complete the study protocol.
variable, with age group (young, old) as the between-sub-
jects factor and the three experimental conditions (No
Adaptation, Visible Adaptation, Non-Visible Adap-
3.1. Comparison between subsamples
tation) as the within-subject factors. Following the ana-
To learn about the differences between both user groups lyses of variance, we computed Bonferroni corrected t-
regarding manual dexterity and technical affinity, we tests to test the differences between the levels of the
conducted the following tests. within-subjects factor, when appropriate.
Manual dexterity – Box and Block Test There was a Range of adaptation A main effect of adaptation type,
significant difference in the manual dexterity scores of F(1, 46) = 5.36, p = .025, h2G = .022, was found in
the young age group (M = 73.7, SD = 7.05) and the old regard to the button-end-size. This suggests that adap-
age group (M = 60.4, SD = 8.78); t(43.9) = −5.78, tation types did in fact significantly differ from each
p < .001, d = 1.67 as measured with the Box and Block other, the button-end-size of the non-visible adaptation
Test. We conducted a simple linear regression to predict being smaller than that of the visible adaptation (see
the number of achieved blocks during the test based on Figure 3). Neither a significant main effect of age
the participant’s age. A significant regression equation group, F(1, 46) = 1.42, p = .24, h2G = .024, nor an inter-
was found, F(1, 46) = 44.4, p < .001, with an R2 of .49. action of age group and adaptation type,
Participant’s predicted number of blocks is equal to F(1, 46) = 0.18, p = .67, h2G = .0008, were found.
 BEHAVIOUR & INFORMATION TECHNOLOGY 7

Figure 3. Adaptation-range results, contrasted by age group Figure 4. Results of the contrasts of age group (young compared
(young, light grey, compared to senior participants, dark grey) to senior participants) and the three adaptation levels. Left: com-
and the two adaptation type levels: visible (left) and non-visible parison between old and young in the no adaptation condition.
(right). The dashed line represents the size of the no adaptation Center: comparing the visible adaptation condition. Right: con-
condition. All error bars represent the SEM. trasting the total hit ratio of young and old participants in the
non-visible condition. All error bars represent the SEM.
Button-end-size ranged for the young group from 31 to
73 points (SD = 10.8) and for the older group from 33 to < .001, h2G = .13. Post-hoc comparisons revealed
111 points (SD = 13.8 ), further suggesting that partici- highly significant differences between all condition
pants in each group differed from each other in regard combinations (p < .001). As can be seen from the
to the amount of adaptation, which they elicited (see results of the statistical analysis and data presented
Figure 3). in Figure 3, spell speed was highest (i.e. Time-per-Cor-
Interaction accuracy Significant main effects of age rect-Letter lowest) in both groups during the visible
group, F(1, 46) = 19.1, p < .001, h2G = .18, and adap- adaptation, followed by the non-visible adaptation
tation type, F(1.1, 50.51) = 184.1, p < .001, h2G = .65 condition. As for the difference between age groups,
on the interaction accuracy were found. Furthermore, the interaction speed results suggest that the difference
a significant interaction of age group and adaptation between older and younger participants was highly sig-
type could be obtained, F(1.1, 50.51) = 28.1, p < .001, nificant, and older participants were generally slower
h2G = .22. Post hoc comparisons revealed a significant than younger. The interaction between age group
difference between visible and non-visible adaptation, and adaptation condition suggests that elderly partici-
p = .031, between no-adaptation and visible adaptation, pants’ interaction speed increased due to the avail-
p < .001 and between no-adaptation and non-visible ability of adaptation more than that of younger
adaptation, p < .001. Results (shown in Figure 3) suggest participants. These could still interact in a timely
that the user’s accuracy was best in the non-visible adap- fashion during the no-adaptation condition and
tation condition, followed by the visible adaptation con- showed significant but smaller improvement due to
dition. Results further show that younger participants adaptation. Moreover, participants of the older group
were generally more accurate than older participants. benefited most from the adaptation conditions, almost
The interaction of the age group and adaptation con- doubling their interaction speed when adaptation was
dition suggests that the older age group profited more available (see Figure 5).
from the availability of adaptation than the young Workload As can be seen in Figure 6, the adap-
group (see Figure 4). tation conditions induced different levels of workload,
Interaction speed There were significant main effects F(1.73, 79.75) = 5.37, p = .006, h2G = .032. Post hoc
of both age group, F(1, 46) = 43.0, p < .001, h2G = .38, comparisons revealed that the visible adaptation con-
and adaptation type, F(1.14, 52.52) = 59.6, p < .001, dition induced significantly less perceived effort than
h2G = .31, on interaction speed. Furthermore, the the no-adaptation condition (p = .015). The non-vis-
analysis revealed a significant interaction of age ible condition showed a tendency for lower subjective
group and adaptation type F(1.14, 52.52) = 20.5, p workload in comparison to the no-adaptation
8 A. SHAHAL ET AL.

and adaptation type, as well as interaction effects of

age group with adaptation type. Older participants
generally rated the scales of Attractiveness, Stimu-
lation, and Novelty higher than the younger group.
The visible and invisible adaptation conditions differed
from the no-adaptation condition mainly on the
Attractiveness, Efficiency, and Stimulation subscales.
However, the results of the post hoc Bonferroni cor-
rections, show that the visible and invisible adaptation
conditions did not significantly differ regarding the
subjective UEQ measurement. Interaction of age
group and adaptation type, which was significant in
the Attractiveness, Perspicuity, Efficiency, Stimulation,
and Novelty subscales suggests that the difference
between the rating of adaptation conditions differed
according to age groups. The difference in usability
Figure 5. Comparisons of interaction speeds (time in seconds between the no-adaptation and the adaptation con-
per correct letter) of the three conditions, contrasted by age ditions was higher for the elderly group than for the
group (young, light grey, compared to senior participants, young group.
dark grey). All error bars represent the SEM.

condition (p = .096). Neither a significant main effect

4. Discussion
of age group, F(1, 46) = 0.06, p = .81, h2G = .0009, 4.1. Range of adaptation
nor an interaction effect, F(1.73, 79.75) = 1.77, p
Overall, results suggest that both age groups took
= .18, h2G = .011, were found. The subjective effort
for each condition generally indicated a relatively advantage of adaptation as implemented in this study.
low effort level across both groups and all conditions This was evident from the comparison of the button-
end-size with the button’s initial size at the beginning
(M = 50.7, SD = 40.9 ). This result could reflect a
floor effect, suggesting that the subjective effort of each condition. Furthermore, in both age groups,
induced through the interaction with the experimental different participants took advantage of the adaptation
to different extents, requiring practically no change of
task was generally lower than what can meaningfully
be captured using the SEA scale. button size. In contrast, others required an adaptation,
User Experience The results of the multiple analyses which was more than 3-times the original button size.
This finding suggests that, although the adaptation
of variance, performed separately for each of the UEQ
algorithm only allowed enlargement of the button size,
subscales, can be seen in Table 1. Findings from all
UEQ scores suggest both main effects of age group the mechanism itself was sensitive to participants’ accu-
racy. It was elicited when accuracy was low, and not eli-
cited when accuracy was sufficient.

4.2. Effects of adaptation on different age groups

Elderly and younger users’ touch interaction differed to
various extents, depending on the observed measure-
ment. The results of the range of adaptation did not
reveal significant differences between age groups.
Although the latter can be regarded as an indicator of
the extent to which participants took advantage of the
benefits of adaptation, it could nevertheless be con-
founded by each age group’s interaction strategy.
While receiving the same instructions, interaction
data, and observations of users’ interaction suggest
Figure 6. Bar plots of the subjective experienced effort for all that younger participants were generally more confi-
conditions and both age groups. Error bars represent the SEM. dent. In contrast, older adults tended to interact slower
 BEHAVIOUR & INFORMATION TECHNOLOGY 9

Table 1. Summary of the multiple analyses of variance, performed for each of the six UEQ subscales. Reported post-hoc analyses were
performed using a Bonferroni correction.
Post-hoc comparisons between
adaptation types
UEQ Subscale Effect F dfs p h2G NA:VA NA:IA VA:IA
Attractiveness Age Group 4.29 1, 46 .043 .065 – – –
Adaptation Type 10.4 1.62, 74.38 ≤.001 .054 .001 .019 n.s.
Age Group × Adaptation Type 3.79 1.62, 74.38 .035 .02 – – –
Perspicuity Age Group ≤1 1, 46 n.s. – – – –
Adaptation Type 9.07 1.29, 59.59 .002 .065 .005 .024 n.s.
Age Group × Adaptation Type 5.59 1.29, 59.59 .014 .041 – – –
Efficiency Age Group 1.16 1, 46 n.s. .018 – – –
Adaptation Type 7.07 1.75, 80.33 .002 .04 .024 .013 n.s.
Age Group × Adaptation Type 3.7 1.75, 80.33 .034 .022 – – –
Dependability Age Group ≤1 1, 46 n.s. – – – –
Adaptation Type 2.28 2, 92 n.s. – – – –
Age Group × Adaptation Type 1.14 2, 92 n.s. – – – –
Stimulation Age Group 8.07 1, 46 .007 .13 – – –
Adaptation Type 9.46 2, 92 ≤.001 .022 ≤.001 .015 n.s.
Age Group × Adaptation Type 3.32 2, 92 .040 .008 – – –
Novelty Age Group 14.1 1, 46 ≤.001 .22 – – –
Adaptation Type 2.77 2, 92 .067 .005 – – –
Age Group × Adaptation Type 8.00 2, 92 ≤.001 .014 – – –
Note: NA = No Adaptation, VA = Visible Adaptation, IA = Invisible Adaptation.

and more cautiously with the system. The speed-accu- stimulating, and novel than younger adults. This differ-
racy tradeoff, in which older users apply speed reduction ence could either imply that the spelling task is indeed
in order to maintain interaction accuracy, has often more appropriate for the elderly group, or, as with the
been discussed in the literature regarding aging (Birren workload measurement, it is the result of different rat-
et al. 2001; Laugwitz, Held, and Schrepp 2008; Craik and ing-tendencies of both age groups. Here too, no prior
Salthouse 2011). Therefore, both age groups could have research could be found to support this notion. Results
taken similar advantage of adaptation while being dri- of the interaction between age-group and adaptation-
ven by two distinct interaction strategies. To further condition imply that older adults rated adaptation
evaluate each age group’s interaction, parameters of higher than no-adaptation. At the same time, younger
interaction accuracy and speed were analysed. Both par- participants tended to be more immune to the size of
ameters showed significant differences between age interactive elements and the benefits of adaptation. In
groups and their interaction with each of the adaptation conclusion, comparison of both age groups has empha-
conditions. As expected, the elderly group generally had sised previously suggested adverse effects of interaction-
a lower interaction accuracy and was also slower in element size on the usability of touch displays for elderly
interaction than the younger group. Both age groups users (e.g. Claypoole, Schroeder, and Mishler (2016)).
took advantage of adaptation and showed better accu- Older users were more prone to size-related problems.
racy and higher typing speed when adaptation was Utilising user-specific adaptation can diminish such
available. However, the interaction of age group and size-related problems and enable older users to reach
adaptation type suggests that the older group was sig- interaction characteristics similar to those of younger
nificantly more prone to size-related difficulties in its adults. Subjective measurements suggest that adaptation
interaction with the touch device. Interestingly, data does not require more effort to be invested in the system
also suggest that once adaptation was available, older interaction. Both groups found adaptation acceptable
adults could reach hit ratios and spell speeds, similar and preferred the interaction with the system when
to those of the younger group (cf. Figure 4). Concerning adaptation was available. Thus, adaptive solutions
the costs of interaction with the spelling task, the subjec- could offer a promising, Design-for-All oriented sol-
tively perceived effort scale (Eilers, Nachreiner, and ution for touch-interfaces, effectively reducing the limit-
Hänecke 1986) did not reveal significant differences ation currently perceived by elderly users in their
between age groups and generally showed low levels interaction with touch devices.
of perceived effort. Future evaluations should consider
using another measure to exclude a possible floor-
4.3. Effects of adaptation visibility
effect. Subjective measures using the UEQ (Leonardi
et al. 2010) suggest that older adults generally perceived Objective measurements of differences between adap-
the interaction with the spelling task as more attractive, tation types revealed several findings. Concerning the
10 A. SHAHAL ET AL.

range of adaptation, the button-end-size measurement differences between the no-adaptation and both adap-
revealed that the final button size was smaller in the tation conditions on the UEQ sub-scales of Attractive-
non-visible adaptation compared to the visible adap- ness, Effectivity, Perspicuity, and Stimulation.
tation for both age groups. This result suggests that However, both adaptation conditions did not signifi-
compared to visible adaptation, the non-visible adap- cantly differ from each other. This suggests that while
tation elicited less change in button size. This could be adaptation, in general, was found advantageous in the
related to the lack of visual change in the target button’s current experimental setting, participants did not prefer
area, requiring higher accuracy from the user than in the either of the adaptation types to the other.
visible adaptation. The interaction accuracy measure-
ment, in which non-visible adaptation proved to be
4.4. Limitations and future research
more accurate than visible adaptation, complements
this finding. It suggests that non-visible adaptation led Owing to our recruitment method (which was primarily
to the least misses of target buttons. Both adaptation performed via email), this study’s results reflect older
conditions led to significantly higher interaction accu- adults’ abilities with basic computer competence
racies when compared with the no-adaptation con- (which might not necessarily represent the general
dition. Although participants elicited less adaptation population of older adults). Nevertheless, our findings
and were more accurate with the non-visible adaptation, are in line with (Hwangbo et al. 2013) who found a sig-
their interaction speed was higher in the visible adap- nificant impact of target spacing on pointing perform-
tation condition. As interaction elements were ran- ance – while sampling from non-smartphone owning
domly spread on the screen, response speed was not population. This emphasises the presented results in
only affected by the ease of physical interaction with general and the advantage of adaptation for elderly
the system, but also from the search for the target but- users in particular. While keeping the interaction area
ton. With increasing button size in the visible adap- of all buttons equal, it is clear that individual letters
tation condition, a decrease in the visual search task have different visual footprints (e.g. ‘I’ vs. ‘M’). Future
difficulty was evident. This is in line with Fitts’ Law research could reduce the risk of a specific letter affect-
(Fitts 1954), suggesting that the motion time (to reach ing the visual search, for example, by implementing a
a target) is partially dependent on the target size. There- frame surrounding each button. Age-related differences
fore, the superiority of the visible adaptation over the in the responses to the SEA scale could have con-
non-visible adaptation could be seen as a byproduct of founded our findings: If the elderly group were to report
the experimental paradigm, in which a visual search different effort ratings than the younger group for the
task affected the response speed. Interestingly, although same perceived effort, the comparison of both groups
it could be assumed that change in target size would would not be valid. Although this possibility exists, it
mostly affect the group, for which the visual task is could not be supported by the reviewed literature. Con-
the hardest (i.e. older adults), this was not the case. cerning the chosen algorithm, as the primary goal of this
The spell speed of both older and younger participants experiment was to study the fundamental aspects of
was similarly affected by the increase in target size. Sub- adaptation and its effect on the interaction of different
jective measurement of the effort revealed a significant age groups with touch devices, it was decided to utilise
difference between the no-adaptation condition and an algorithm that allowed only one direction of adap-
the visible adaptation condition, but not between the tation. Although this decision was beneficial in many
no-adaptation and non-visible adaptation or between aspects, it is clear that the choice of the adaptation algor-
the two adaptation conditions. This suggests that par- ithm restricts our findings’ generalisability. In order to
ticipants of both age-groups found the visible adap- increase the relevance of this study’s findings, testing
tation to require less effort than the no-adaptation and our hypotheses with further algorithms, which have
non-visible adaptation conditions. As with the above- proven successful in previous studies (e.g. Saffer
mentioned implications of target size on interaction (2008); Buschek, Rogers, and Murray-Smith (2013)),
speed, results of the subjectively perceived effort could would be advised.
further reflect an effect of the visual search. This should
not reduce the importance of the finding. Visible adap-
4.5. Practical implications
tation in this specific experimental setting required less
effort than non-visible adaptation. Instead, this calls for The results of this study offer several recommendations
further investigation in which the effect of the visual for the design of touch displays in general, and the
search task will be controlled. Subjective usability design of touch displays for elderly users in particular.
measurements using the UEQ scale revealed significant In line with previous findings (Jin, Plocher, and Kiff
 BEHAVIOUR & INFORMATION TECHNOLOGY 11

2007; Kobayashi et al. 2011; Hwangbo et al. 2013), our Considering these results, it seems that while the
results suggest that the elderly user group is indeed implementation of visible or non-visible adaptation is
more affected by problems related to the size of inter- favourable compared to a system with no adaptation,
action elements than younger users. An emphasis on the matter of visibility can be specifically designed to
the size of interaction elements is, therefore, of particu- fit the system’s characteristics and goals. That being
lar interest when designing touch interaction for elderly said, both types have natural limitations in practical
users. Adapting the size of interactive elements to the implementations: visible adaptation might cause
needs of the specific user improved the interaction of dynamic interface changes up to distortions, potentially
both tested age groups with the touch device. This was causing design, structure, and space issues. Likewise,
especially helpful for the elderly group. The proneness invisible adaptation is limited to interfaces that provide
of elderly users to problems related to the size of inter- enough space around interaction elements not to over-
action elements could be alleviated through adaptation, lap their interaction-sensitive boundaries. Our test setup
allowing elderly users to reach interaction accuracy and with relatively few buttons on a 9.7-inch display offered
speed comparable to those of younger users. Thus, the enough space to increase hit-boxes around icons with-
implementation of adaptation in systems that are to out introducing overlapping. This technique might be
be used by elderly users, among others, could generally challenging to adapt for interfaces communicating a
be recommended. The extent to which adaptation was dense amount of information or complex menu struc-
elicited varied between participants of both age groups, tures. A possible inspiration for dealing with this issue
suggesting that custom-fit adaptation can serve users could be borrowed from responsive web design: here,
experiencing more difficulty in tapping a required different break-points define different layouts, depend-
area. At the same time, this does not hinder the inter- ing on the available window size. This concept could
action of users interacting effortlessly with an interface be extended to depend on viewport width and the avail-
and thus do not need size adaptation. Therefore, it able space usable for button size changes. If UI elements
seems that a custom-fit adaptation of interaction overlap after size adjustments, the layout could tran-
elements is beneficial in fulfilling the goals of DfA, sition into a trimmed down version, offering more
again allowing a broader spectrum of users to take space for adaptive button sizes. However, this approach
advantage of a given device. Furthermore, instead of is also limited to a reasonable maximum of different lay-
using touch devices specifically designed for the elderly outs, button size, and depending on available screen size
population, adaptation could allow older users to use and content.
conventional devices, thus avoiding possible stigmatis-
ation and being able to enjoy mainstream products.
4.6. Conclusion
Subjective measures suggest that while adaptation was
experienced as advantageous by both groups, no nega- To gain insight into the relevance of adaptive touch dis-
tive consequences could be identified when comparing plays for elderly users, different adaptive solutions for
the system in its adaptive configuration, with the non- the size of interaction elements on touch screens were
adaptive configuration. Thus, subjective results under- implemented and tested in a laboratory setting with
line the recommendation, as mentioned above, in younger and older adults. Visible and invisible adap-
favour of adaptation in general. As to the comparison tation conditions were compared with each other and
of adaptation types, objective measures suggest that with a control condition of no-adaptation. A counterba-
non-visible adaptation allows for more accurate inter- lanced design using both objective and subjective
action with the system than visible adaptation while eli- measures was employed. Results suggest that elderly
citing less change. This allows us to differentiate users suffer more from an inappropriate size of inter-
between user-specific adaptation and such that is action elements than younger adults and that the inter-
more context-based. For example, invisible adaptation action of elderly users with touch displays can be
might be more suitable for tasks requiring more precise substantially improved through the implementation of
control (e.g. less error-tolerant). In contrast, visible a user-specific adaptation of the graphical user interface.
adaptation might be especially interesting for reaction A comparison of visible and non-visible adaptation
time-critical tasks. At the same time, visible adaptation suggests that non-visible adaptation is more accurate
allows for quicker interaction with the system. This and elicits less need for adaptation. In contrast, visible
result may have been, as previously suggested, confined adaptation enables a higher interaction speed with the
by the effects of the visual search, and requires further system. In a world where touch interaction gained
evaluation. Regarding subjective measures, adaptation great popularity as the Human-Machine interaction
types do not significantly differ from each other. method of choice, user-specific adaptation could enable
12 A. SHAHAL ET AL.

a broader and more heterogenic user group to enjoy Chung, M. K., D. Kim, S. Na, and D. Lee. 2010. “Usability
touch devices and allow designers of Information and Evaluation of Numeric Entry Tasks on Keypad Type and
Communication Technologies to advance in the direc- Age.” International Journal of Industrial Ergonomics 40
(1): 97–105.
tion of Design for All. Claypoole, V. L., B. L. Schroeder, and A. D. Mishler. 2016.
“Keeping in Touch: Tactile Interface Design for Older
Users.” Ergonomics in Design 24 (1): 18–24.
Acknowledgments Craik, F. I., and T. A. Salthouse. 2011. The Handbook of Aging
and Cognition. Psychology Press.
We would like to thank the Senior Research Group (SRG) at
Darroch, I., J. Goodman, S. Brewster, and P. Gray. 2005. “The
Technische Universität Berlin for their support and many
Effect of Age and Font Size on Reading Text on Handheld
helpful comments.
Computers.” In IFIP Conference on Human-Computer
Interaction, Rome, 253–266. Springer.
Dunn, W., J. W. Griffith, D. Sabata, M. T. Morrison, J. C.
Disclosure statement MacDermid, A. Darragh, R. Schaaf, B. Dudgeon, L. T.
No potential conflict of interest was reported by the author(s). Connor, L. Carey, and J. Tanquary. 2015. “Measuring Change
in Somatosensation Across the Lifespan.” American Journal of
Occupational Therapy 69 (3): 6903290020p1–6903290020p9.
Eilers, K., F. Nachreiner, and K. Hänecke. 1986. “Entwicklung
ORCID Und Überprüfung Einer Skala Zur Erfassung Subjektiv
Avner Shahal http://orcid.org/0000-0002-5406-9148 Erlebter Anstrengung.” Zeitschrift Für Arbeitswissenschaft
Robert P. Spang http://orcid.org/0000-0001-6580-9060 4: 214–224.
Michael Minge http://orcid.org/0000-0003-1930-7064 Fitts, P. M. 1954. “The Information Capacity of the Human
Carola Trahms http://orcid.org/0000-0003-1055-1896 Motor System in Controlling the Amplitude of Movement.”
Jan-Niklas Voigt-Antons http://orcid.org/0000-0002-2786- Journal of Experimental Psychology 47 (6): 381–391.
9262 Gena, C. 2005. “Methods and Techniques for the Evaluation
of User-Adaptive Systems.” The Knowledge Engineering
Review 20 (1): 1–37.
Goldstein, M., R. Book, G. Alsiö, and S. Tessa. 1999. “Non-
References
Keyboard Qwerty Touch Typing: A Portable Input
Antons, J., J. O’Sullivan, S. Arndt, P. Gellert, J. Nordheim, S. Interface for the Mobile User.” In Proceedings of the
Mller, and A. Kuhlmey. 2016. “Pflegetab: Enhancing SIGCHI Conference on Human Factors in Computing
Quality of Life using a Psychosocial Internet-Based Systems, Pittsburgh Pennsylvania, 32–39. ACM.
Intervention for Residential Dementia Care.” In Goodman, J., G. Venolia, K. Steury, and C. Parker. 2002.
Proceedings of International Society for Research on “Language Modeling for Soft Keyboards.” In Proceedings
Internet Interventions, International Society for Research of the 7th International Conference on Intelligent User
on Internet Interventions, 1–1. Elsevier B.V. Interfaces, San Francisco, California, 194–195. ACM.
Arnett, J. J. 2008. “The Neglected 95%: Why American Gunawardana, A., T. Paek, and C. Meek. 2010. “Usability
Psychology Needs to Become Less American.” American Guided Key-Target Resizing for Soft Keyboards.” In
Psychologist 63 (7): 602–614. Proceedings of the 15th International Conference on
Barnard, Y., M. D. Bradley, F. Hodgson, and A. D. Lloyd. 2013. Intelligent User Interfaces, Hong Kong, 111–118. ACM.
“Learning to Use New Technologies by Older Adults: Hanel, P. H., and K. C. Vione. 2016. “Do Student Samples
Perceived Difficulties, Experimentation Behaviour and Provide An Accurate Estimate of the General Public?”
Usability.” Computers in Human Behavior 29 (4): 1715–1724. PloS One 11 (12): e0168354.
Birren, J. E., K. W. Schaie, R. P. Abeles, M. Gatz, and T. A. Hertzum, M., and K. Hornbæk. 2010. “How Age Affects
Salthouse. 2001. Handbook of the Psychology of Aging. Pointing with Mouse and Touchpad: A Comparison of
Vol. 2. Gulf Professional Publishing. Young, Adult, and Elderly Users.” Intl Journal of
Bühler, C., and C. Stephanidis. 2004. “European Co- Human–Computer Interaction 26 (7): 703–734.
Operation Activities Promoting Design for All in Hwangbo, H., S. H. Yoon, B. S. Jin, Y. S. Han, and Y. G. Ji.
Information Society Technologies.” In International 2013. “A Study of Pointing Performance of Elderly Users
Conference on Computers for Handicapped Persons, Paris, on Smartphones.” International Journal of Human-
France, July 7-9, 80–87. Springer. Computer Interaction 29 (9): 604–618.
Buschek, D., S. Rogers, and R. Murray-Smith. 2013. “User- Jin, Z. X., T. Plocher, and L. Kiff. 2007. “Touch Screen User
Specific Touch Models in a Cross-Device Context.” In Interfaces for Older Adults: Button Size and Spacing.” In
Proceedings of the 15th International Conference on International Conference on Universal Access in Human-
Human-Computer Interaction with Mobile Devices and Computer Interaction, Beijing, 933–941. Springer.
Services, Munich, 382–391. ACM. Karrer, K., C. Glaser, C. Clemens, and C. Bruder. 2009.
Choi, N. G., and D. M. DiNitto. 2013. “The Digital Divide “Technikaffinität Erfassen–Der Fragebogen Ta-eg.” Der
Among Low-Income Homebound Older Adults: Internet Mensch Im Mittelpunkt Technischer Systeme 8: 196–201.
Use Patterns, Ehealth Literacy, and Attitudes Toward Ketcham, C. J., and G. E. Stelmach. 2004. “Movement Control
Computer/Internet Use.” Journal of Medical Internet in the Older Adult.” In Technology for Adaptive Aging.
Research 15 (5): e93. National Academies Press (US).
 BEHAVIOUR & INFORMATION TECHNOLOGY 13

Kobayashi, M., A. Hiyama, T. Miura, C. Asakawa, M. Hirose, Saffer, D. 2008. Designing Gestural Interfaces: Touchscreens
and T. Ifukube. 2011. “Elderly User Evaluation of Mobile and Interactive Devices. O’Reilly Media Inc.
Touchscreen Interactions.” In IFIP Conference on Smith, M. W., J. Sharit, and S. J. Czaja. 1999. “Aging, Motor
Human-Computer Interaction, Lisbon, 83–99. Springer. Control, and the Performance of Computer Mouse
Laugwitz, B., T. Held, and M. Schrepp. 2008. “Construction Tasks.” Human Factors 41 (3): 389–396.
and Evaluation of a User Experience Questionnaire.” In Stephanidis, C. 2000. User Interfaces for All: Concepts,
Symposium of the Austrian HCI and Usability Engineering Methods, and Tools. CRC Press.
Group, Graz, 63–76. Springer. Stößel, C., H. Wandke, and L. Blessing. 2010. Gestural
Lavie, T., and J. Meyer. 2010. “Benefits and Costs of Adaptive Interfaces for Elderly Users: Help or Hindrance?” In
User Interfaces.” International Journal of Human- Gesture in Embodied Communication and Human-
Computer Studies 68 (8): 508–524. Computer Interaction, edited by S. Kopp, I. Wachsmuth,
Leonardi, C., A. Albertini, F. Pianesi, and M. Zancanaro. 2010. 269–280. Berlin, Heidelberg: Springer Berlin Heidelberg.
“An Exploratory Study of a Touch-Based Gestural Interface Walker, N., D. A. Philbin, and A. D. Fisk. 1997. “Age-Related
for Elderly.” In Proceedings of the 6th Nordic Conference on Differences in Movement Control: Adjusting
Human-Computer Interaction: Extending Boundaries, Submovement Structure to Optimize Performance.” The
Reykjavik, 845–850. ACM. Journals of Gerontology Series B: Psychological Sciences
Mathiowetz, V., G. Volland, N. Kashman, and K. Weber. and Social Sciences 52 (1): P40–P53.
1985. “Adult Norms for the Box and Block Test of Watson, D., M. Hancock, R. L. Mandryk, and M. Birk. 2013.
Manual Dexterity.” American Journal of Occupational “Deconstructing the Touch Experience.” In Proceedings of
Therapy 39 (6): 386–391. the 2013 ACM International Conference on Interactive
Mock, P., J. Jaszkowic, J. Edelmann, Y. Kammerer, A. Schilling, Tabletops and Surfaces, St Andrews, Scotland, 199–208.
and W. Rosenstiel. 2015. “Keyvaluate: A Framework for ACM.
Rapid Evaluation of Adaptive Keyboards on Wigdor, D., and D. Wixon. 2011. Brave NUI World: Designing
Touchscreens.” In Proceedings of the Annual Meeting of Natural User Interfaces for Touch and Gesture. Elsevier.
the Australian Special Interest Group for Computer Human Wright, P., C. Bartram, N. Rogers, H. Emslie, J. Evans, B.
Interaction, Parkville VIC Australia, 361–369. ACM. Wilson, and S. Belt. 2000. “Text Entry on Handheld
Naumann, A. B., I. Wechsung, and J. Hurtienne. 2010. Computers by Older Users.” Ergonomics 43 (6): 702–716.
“Multimodal Interaction: A Suitable Strategy for Ziefle, M. 2001. “Aging, Visual Performance and Eyestrain in
Including Older Users?” Interacting with Computers 22 Different Screen Technologies.” In Proceedings of the
(6): 465–474. Human Factors and Ergonomics Society Annual Meeting,
Nordheim, J., J. O’Sullivan, P. Gellert, S. Arndt, A. Kuhlmey, Vol. 45, 262–266. Los Angeles, CA: SAGE Publications.
and J. Antons. 2017. “Tablet-Based Intervention for Ziefle, M., and S. Bay. 2005. “How Older Adults Meet Complexity:
Nursing Home Residents with Dementia: A Cluster Aging Effects on the Usability of Different Mobile Phones.”
RCT.” Innovation in Aging 1 (1): 1370–1370. Behaviour & Information Technology 24 (5): 375–389.