Effects of Cognitive Load Variation on Anthropomorphism During a

Cooperative Human-Robot Pick-and-Place Task

Mohamed Cherif Rais

, Barbara Kühnlenz

2 a

and Kolja Kühnlenz

1 b

Robotics Research Lab, Department of Electrical Engineering and Computer Science, Coburg University of Applied

Sciences and Arts, D-96450 Coburg, Germany

Business Psychology and Human-Machine Interaction, Department of Economics, Ansbach University of Applied

Sciences, D-91522 Ansbach, Germany

Keywords: Human-Robot Interaction, Anthropomorphism, Cognitive Load.

Abstract: This paper investigates anthropomorphism of a robot arm during a cooperative human-robot pick-and-place

task, while varying cognitive load of test persons. Test persons are required to repeatedly provide a Lego brick

for the robot by alternatingly putting it onto one of two trays. The robot then picks it up and puts it in front of

the test person again. Cognitive load is varied by whether or not an initially given 8-digit number has to be

remembered by the test person. Dimensions of anthropomorphism are acquired using the HRIES

questionnaire and cognitive load is acquired using two state-of-the-art questionnaires. Results show a

significant correlation of perceived sociability and animacy on mental demand and cognitive load, but only

in the high load condition. It is suggested, that cognitive load should be considered during cooperative task

design because resulting variations of anthropomorphism may impact cooperative task performance.

1 INTRODUCTION

Human-like form and function can increasingly be

noted in many robotic products. Especially, in the

field of industrial robots, this is an interesting trend

(e.g. Sawyer (ReThink Robotics), YUMI (ABB),

etc.). Intuitiveness of how to interact with the robot is

one reason (Mayer et al., 2012), e.g. in terms of

predictability and legibility of actions (Brecher et al.,

2013; Kirsch et al., 2010). It is further assumed, that

still human workers will be required in many

production environments, e.g. for particular tasks in

the area of assembly (Faber et al., 2015), where

flexibility, cognitive abilities and sensorimotor skills

of the collaborating human is complemented by

endurance and power of the robot (Shen and Reinhart,

2013). Further, the ongoing trend towards small batch

sizes is difficult to be accounted for by many SMEs

in terms of complete automation of production lines,

which may be uneconomical (Bley et al., 2004). In

this context, Challenges and potential contribution of

collaboration of human and robot in manufacturing

https://orcid.org/0000-0002-9214-1032

https://orcid.org/0000-0003-1511-9381

are recognized especially for high-wage countries

(Petruck et al., 2016).

The effect of anthropomorphism, i.e. attributing

human-like qualities to an object, is therefore

important to be accounted for in human-robot

interactive collaboration as it is known to impact

collaborative task-performance as well as user

experience. However, it is also known, that cognitive

load has a significant effect on anthropomorphism

(Spatola and Chaminade, 2022), whereas higher

cognitive load leads to increased levels of

anthropomorphism as two main processes co-exist in

the human brain concerning social and physical

cognition with the social pathway being the default of

cognitive process-sing, e.g. (Mars et al., 2012). Social

cognition is further related to attributing socially

relevant parameters, e.g. anthropomorphism. This

default pathway, however, may be inhibited by

processes, which are concerned with physical tasks

(Darlow and Sloman, 2010; Evans and Stanovich,

2013). So far, however, this effect is not yet

investigated within physical human-robot interaction

contexts as in related work, only robot videos are

314

Rais, M., Kühnlenz, B. and Kühnlenz, K.

Effects of Cognitive Load Variation on Anthropomorphism During a Cooperative Human-Robot Pick-and-Place Task.

DOI: 10.5220/0013057700003822

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 21st International Conference on Informatics in Control, Automation and Robotics (ICINCO 2024) - Volume 2, pages 314-318

ISBN: 978-989-758-717-7; ISSN: 2184-2809

displayed without any physical task to be

accomplished by the participants (Spatola and

Chaminade, 2022).

Furthermore, also adverse effects are present,

which result from human-like design of robots. In this

context, it is known from human-human interaction,

that hand movements may deviate from task

movements in the direction of the hand movements of

a counterpart, which is also found to be present in

human-robot interaction (Kilner et al., 2003;

Chaminade et al., 2005; Kühnlenz and Kühnlenz,

2020; Kupferberg et al., 2011) and which is assumed

to be caused by activation of the mirror neuron system

and the implication of a tendency to imitate a human

counterpart to some extent. Also here, the effect of

anthropomorphism of robots may be highly relevant.

This paper investigates, in how far cognitive load

affects anthropomorphism during a real cooperative

human and robot pick-and-place task. An object

placement scenario is chosen, where robot and human

are positioned in front of and facing each other and

test persons have to repeatedly place an object

alternatingly on a left and right tray to be picked up

by the robot, which puts it back to an initial tray.

The remainder of the paper is organized as follows:

In Section 2, the hypothesis and study design are

presented; results are shown in Section 3 and discussed

in Section 4; conclusions are given in Section 5.

2 HYPOTHESES

As outlined, a significant dependency of

anthropomorphism dimensions on cognitive load is

expected as well as a positive association of both

during a physical human-robot cooperation task. So,

the main hypotheses are:

H1: Anthropomorphism of a robot depends on

cognitive load during physical cooperation of

human and robot.

H2: Anthropomorphism and cognitive load are

correlated positively during physical

cooperation of human and robot.

3 METHODOLOGY

In this paper, we exemplarily explore the effect of

cognitive load of a test person on anthropomorphism

of a cooperating robot. A cooperative pick-and-place

task was chosen, where the test person has to place a

Lego brick on one of two trays in an alternating way,

whereafter a Panda robot arm (Franka Emika)

repeatedly put the Lego brick back to an initial

position, where it is picked up then again by the test

person. Two randomized conditions (counterbalanced,

low load vs. high load) are realized as independent

factor variable in a within-subjects design, which differ

in whether an additional mental task has to be

performed constituting a certain additional cognitive

load for the test person.

Induction methodologies for cognitive load are

manifold, whereas three main techniques are

commonly used: number memorization, visual

pattern and auditory recall tasks, and time pressure

(Deck et al., 2021). Due to practical reasons in the

context of the chosen scenario, we decided for

number memorization, where the test person is shown

an 8-digit number, which they are told to memorize

during the task and write down afterwards.

Figure 1: Experimental set-up.

Initially, test persons are briefed towards the

experimental procedure, but not the hypotheses, and

informed consent is obtained. The persons then have

to complete a pre-questionnaire acquiring

demographic data (age, gender), previous experience

with robots, field of study, and left-/right-handedness.

Afterwards the cooperation task is performed in each

of the two cognitive load conditions in randomized

way, whereas a post-questionnaire is completed after

each condition. Finally, a de-briefing is conducted.

The post-questionnaire acquires dimensions of

perceived anthropomorphism as well as task-load

based on the validated Human-robot Interaction

Evaluation Scale HRIES (4 constructs (sociability,

tray 1

tray 2

initial position

Effects of Cognitive Load Variation on Anthropomorphism During a Cooperative Human-Robot Pick-and-Place Task

315

animacy, agency, disturbance) of 4 items each, 7-

point Likert scale) (Spatola et al., 2021), and NASA

TLX (5-point Likert-scale), respectively. In the latter,

the sub-scale ‘mental demand’ is assumed to be

correlated with cognitive task load. In addition,

another construct for cognitive load measurement is

used in order to be comparable to (Spatola and

Chaminade, 2022), which includes the two questions

“Regarding the task you performed, the task was very

complex” and “Regarding the task you performed,

you provided a very high mental effort to solve it”.

For reasons of clarity, this measure is abbreviated

‘CL’ in the following.

The procedure for one test person is shown in

Table 1.

Table 1: Procedure for one test person.

Task

Du duration

comments

briefing

introduction (context, procedure, tests)

PRE-quest.

~2 min

demographic data, previous experience

set-up

place test person

adjust

~1 min

let person adjust to setting

cond. 1

~1 min

object placement task in condition 1

POST-quest.

~2 min

NASA TLX, CL, HRIES

cond. 2

~1 min

object placement task in condition 2

POST-quest.

~2 min

NASA TLX, CL, HRIES

de-briefing

4 RESULTS

Results are obtained from 32 test persons (age

between 20 and 42 years, M = 25.3y, SD = 5.8y) and

64 trials (32x2 conditions in a within-subjects

design). All participants are right-handed.

Analyses of covariance (ANCOVA) are conducted

with the HRIES constructs of anthropomorphism

dimensions sociability, animacy, agency, and

disturbance as dependent variables and NASA TLX

and cognitive load items, respectively, constructs as

covariates. Q-Q plots of standardized residuals show

only slightly skewed data with few outliers, so it is

assumed, that ANCOVA results are sufficiently

reliable, in order to obtain conclusive results.

In the high cognitive load condition, ANCOVA

results of sociability with respect to NASA TLX

mental demand and cognitive load construct show a

significant dependency (see Table 2) with a positive

association of both (see Fig. 2). For the remaining

NASA TLX items, no significant effect is observed.

Similarly, ANCOVA results of animacy with

respect to cognitive load construct show a significant

dependency (see Table 3) with a positive association

of both (see Fig. 3). For the NASA TLX items,

however, no significant effect is observed.

For the remaining, HRIES constructs of

anthropomorphism (agency and disturbance) no

significant effects are obtained.

In the low cognitive load condition, no significant

effects are observed.

Table 2: ANCOVA results in high cognitive load

condition; sociability with respect to NASA TLX mental

demand and cognitive load construct (CL).

Cases

Sum of

Squares

Mean

Square

η²

17.659

1.000

17.659

11.262

0.002

0.280

mental

demand

7.886

1.000

7.886

5.029

0.033

0.148

Residual

45.475

29.000

1.568

Note. Type III Sum of Squares

Figure 2: Regression of sociability and cognitive load;

standard error.

Table 3: ANCOVA results in high cognitive load

condition; animacy with respect to NASA TLX mental

demand and cognitive load construct (CL).

Cases

Sum of

Squares

Mean

Square

η²

7.076

1.000

7.076

4.625

0.040

0.138

mental

demand

1.914

1.000

1.914

1.251

0.273

0.041

Residual

44.374

29.000

1.530

Note. Type III Sum of Squares

Table 4: Wilcoxon test results of cognitive load construct

(CL) in the low- (CL(L)) and high-load (CL(H))

conditions.

Rank-Biserial

Correlation

CL(L)

CL(H)

32.500

0.010

-0.658

ICINCO 2024 - 21st International Conference on Informatics in Control, Automation and Robotics

316

Figure 3: Regression of animacy and cognitive load;

standard error.

Figure 4: Q-Q plot; ANCOVA of sociability dependend on

cognitive load (CL).

Figure 5: Q-Q plot; ANCOVA of animacy dependend on

cognitive load (CL).

With respect to checking, whether a different

cognitive load is induced in the two conditions a

Wilcoxon signed-rank test is conducted showing a

significant difference of cognitive load with a strong

effect (see Table 4 and Figure 6).

Figure 6: Cognitive load (CL) in low- (CL(L)) and high-

load (CL(H)) conditions.

Finally, in order to check for the association of the

NASA TLX mental demand scale and cognitive load

(CL) construct scale, a correlation analysis shows a

moderate positive association (p = 0.006, r = 0.48).

5 DISCUSSION

The expected significant positive association between

dimensions of anthropomorphism and cognitive load

could be confirmed with respect to constructs of

sociability and animacy, supporting H1 and H2.

However, this effect is limited to the high cognitive

load condition. In the low cognitive load condition,

such an association is not observed. This

phenomenon could be grounded in a larger variance

in the low cognitive load condition compared to the

high cognitive load condition due to no particular

cognitive load induction is performed in the former.

It would be interesting to investigate the effects using

cognitive load induction in both conditions on

different levels of difficulty.

Effects are of small to moderate strengths (see

Tables 2 and 3), which could be related to the induced

cognitive load differences are also relatively small

(see Table 4 and Figure 6). Stronger effects might be

achieved with larger induced load differences.

Insignificant results regarding agency could be

related to the construct being less suited to evaluate a

robot arm as it refers to perceiving the robot as an

agent as a person having its own beliefs, thoughts and

cognitive load (CL)

CL(L) CL(H)

Effects of Cognitive Load Variation on Anthropomorphism During a Cooperative Human-Robot Pick-and-Place Task

317

intents (Spatola and Chaminade, 2022). Regarding

the disturbance construct, the authors themselves

argue, that the scale is more ambiguous than the

others, which could result in larger variance in the

scenario at hand.

Limitations are given by the moderate sample

size and single interaction task and robot design.

More test persons and different interaction scenarios

may contribute to increase the effects and generalize

the findings.

6 CONCLUSIONS

In the presented human-robot cooperative pick-and-

place study, induction of cognitive load contributed

to a significant positive association of the

anthropomorphism dimensions sociability and

animacy with experienced mental demand and

cognitive load of the test persons. Variable strength

of anthropomorphism due to cognitive load variation

may influence cooperative task performance, which

is generally dependent on the level of perceived

human-likeness of a robot in manifold ways and thus

may impact human-robot interaction quality in a

variety of application scenarios.

Future work will target additional aspects, which

might result from different levels of

anthropomorphism due to cognitive load variation in

the context of safety and trust.

ACKNOWLEDGEMENTS

This paper is supported in part by DAAD.

REFERENCES

Mayer, M., Odenthal, B., Ewert, D., Kempf, T., Behnen, D.,

Bücher, C., Kuz, S., Hauck, E., Kausch, B., Schilberg,

D., Herfs, W., Schlick, C., Jeschke, S. and Brecher, C.

(2012). “Self-optimising assembly systems based on

cognitive technologies,” in Integrative Production

Technology for High-Wage Countries, C. Brecher, Ed.

Springer, pp. 877–990.

Brecher, C., Müller, S., Kuz, S. and Lohse, W. (2013).

“Towards anthropomorphic movements for industrial

robots, likeability, perceived intelli- gence, and

perceived safety of robots,” Proc. 4th Int. Conf. Digit.

Hum. Modeling and Applic in Health, Saf., Erg., and

Risk Man.

Kirsch, A., Kruse, T., Sisbot, A., Alami, R., Lawitzky, M.,

Brscic ́, D., Hirche, S., Basili, P. and Glasauer, S.

(2010). “Plan-based control of joint human-robot

activities,” KI, 24(3):223–231.

Faber, M., Bützler, J. and Schlick, C. (2015). “Human-robot

cooperation in future production systems: Analysis of

requirements for designing an ergonomic work

system,” Procedia Manufacturing, vol. 3, pp. 510– 517.

Shen, Y. and Reinhart, G. (2013). “Safe assembly motion -

a novel approach for applying human-robot co-

operation in hybrid assembly systems,” in Proc. IEEE

ICMA.

Bley, H., Reinhart, G., Seliger, G., Bernardi, M. and Korne,

T. (2004). “Appropriate human involvement in

assembly and disassembly,” CIRP Annals - Man. Tech.,

Elsevier,53:487–509,2004.

Petruck, H., Kuz, S., Mertens, A. and Schlick, C. (2016).

Increasing Safety in Human-Robot Collaboration by

Using Anthropomorphic Speed Profiles of Robot

Movements.

Spatola, N. and Chaminade, T. (2022). Cognitive load

increases anthropomorphism of humanoid robot. The

automatic path of anthropomorphism, International

Journal of Human-Computer Studies, Volume 167.

Mars, R. B., Neubert, F. X., Noonan, M. A. P., Sallet, J.,

Toni, I., Rushworth, M. F. S. (2012). On the

relationship between the “default mode network” and

the “social brain.” Front. Hum. Neurosci., 6, pp. 1-9.

Darlow, A. L., Sloman, S. A. (2010). Two systems of

reasoning: architecture and relation to emotion. Wiley

Interdisciplinary Reviews: Cognitive Science, 1,

pp. 382-392.

Evans, J. S. B. T., Stanovich, K. E. (2013). Dual-process

theories of higher cognition: advancing the debate.

Perspect. Psycholog. Sci., 8 (3), pp. 223-241.

Kilner, J., Paulignan, Y. and Blakemore, S. (2003). “An

interference effect of observed biological movement on

action,” Current Biology, vol. 13, no. 6, pp. 522–525.

Chaminade, T., Franklin, D., Oztop, E. and Cheng, G.

(2005). “Motor interference between humans and

humanoid robots: Effect of biological and artificial

motion,” Proc. IEEE ICDL.

Kühnlenz, K. and Kühnlenz, B. (2020). “Motor interference

of incongruent motions increases workload in close

HRI,” Advanced Robotics, vol. 34, no. 6, p. 400–406.

Kupferberg, A., Glasauer, S., Huber, M., Rickert, M.,

Knoll, A. and Brandt, T. (2011). “Biological movement

increases acceptance of humanoid robots as human

partners in motor interaction,” AI & Society, vol. 26, no.

4, pp. 339–345.

Spatola, N., Kühnlenz, B., Cheng, G. (2021). Perception

and evaluation in human–robot interaction: The

human–robot interaction evaluation scale (HRIES)—a

multicomponent approach of anthropomorphism.

International Journal of Social Robotics, 13(7), pp.

1517-1539.

Deck, C., Jahedi, S. and Sheremeta, R. (2021). On the

consistency of cognitive load, European Economic

Review, Volume 134.

ICINCO 2024 - 21st International Conference on Informatics in Control, Automation and Robotics

318