Analysis of the Relationship between Subjective Difﬁculty of a Task and

the Efforts Put into It using Biometric Information

Katsuko T. Nakahira

, Munenori Harada and Muneo Kitajima

Nagaoka University of Technology, Nagaoka, Niigata, Japan

Keywords:

Biometric information, Difﬁculty of a Task, Pupil Diameter, Learning Motivation.

Abstract:

This paper proposes a novel method for analyzing the relationships between the difﬁculty of tasks and the

effort of learners to accomplish them using biometric information. The biometric information we adopted

was as follows: 1) pupil diameter variation for estimating subjective task difﬁculty, and 2) eye movements

indicative of answer selection times for assessing subjective efforts and strategies to solve the problems. The

data used in this study are eye movement data obtained in a different study for studying brain activities during

arithmetic calculations in terms of electroencephalography (EEG) data (Suzuki et al., 2021). This study re-

analyzed the eye movement data by introducing the following two variables: 1) the duration times in the

characteristic areas for solving the tasks to understand how the participants strategically retrieved the task

information, and 2) the changes in the sizes of pupil diameter to understand the levels of engagement of

the participants while solving the tasks. This study suggests that the relationships found in these variables

should characterize the participants’ learning attitudes and could be related to conﬁdence and satisfaction

in the attention-relevance-conﬁdence-satisfaction (ARCS) model, indicating the possibility of applying the

results to educational systems.

1 INTRODUCTION

Since the onset of the Coronavirus-2019 pandemic,

opportunities for online learning centered on e-

learning have increased from the perspectives of so-

cial distance and avoidance of crowds. This is also

the case in the context of higher education institu-

tions. Additionally, issues related to learning attitudes

such as the quality of learning and the level on moti-

vation have emerged. The intensive e-learning that is

currently practiced, makes it difﬁcult to motivate stu-

dents to learn and keep the quality of learning at high

levels by the timely and appropriate interventions of

the teachers.

One of the reasons for the difﬁculty might be the

discrepancy between the difﬁculty level of the tasks

given to the learner and the learner’s learning abil-

ity. In order for a learner to stay motivated to learn

a subject, it is crucial to have the learner engage in

the task, which has to be at the right level of difﬁ-

culty for the learner. In other words, the subjective

task difﬁculty should be at the right level; when it is

https://orcid.org/0000-0001-9370-8443

https://orcid.org/0000-0002-0310-2796

too easy, the learner would consider the time spent for

solving the task wasteful. On the other hand, if it is

too hard, the learner would give up without gaining

any knowledge from the learning. Learning through

the right task would provide the learner with the sat-

isfaction of task-accomplishment. Even if the learner

fails to solve the problem, s/he would gain useful in-

formation after understanding the correct answer by

gaining knowledge on solving the problem correctly.

In this paper, as a method to solve this discrep-

ancy, we examine the possibility of estimating the ap-

propriateness of the difﬁculty level of the presented

task from the degree of cognitive load at the time of

task execution, using the task of mental arithmetic ad-

dition and multiplication as an example task. In order

to estimate the degree of matching of the difﬁculty

level of a given task with the ability of participants,

this paper examines the utility of biometric informa-

tion such as eye movement and pupil diameter during

calculation.

This paper is organized as follows. Section 2

describes the model of assessment of task difﬁculty

and efforts in solving tasks via biometric information.

Section 3 outlines the previous experiment whose

data we use for this paper. Section 4 describes the

Nakahira, K., Harada, M. and Kitajima, M.

Analysis of the Relationship between Subjective Difﬁculty of a Task and the Efforts Put into It using Biometric Information.

DOI: 10.5220/0010906800003124

In Proceedings of the 17th Inter national Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2022) - Volume 2: HUCAPP, pages

241-248

ISBN: 978-989-758-555-5; ISSN: 2184-4321

241

method for analyzing biometric information. Section

5 describes the results of the analysis. Section 6 dis-

cusses the obtained results from the perspective of ap-

plying them for effective education.

2 SUBJECTIVE TASK

DIFFICULTY AND EFFORTS IN

SOLVING TASKS

This section commences with a description of the

method for estimating the degree of subjective dif-

ﬁculty of tasks by using biometric information in

Section 2.1. It is assumed that the reaction of the

learner at the very moment when s/he observes the

task should be critical to estimating his/her motivation

to engage in the task, and not after s/he spends some

amount of time solving the task. The biometric infor-

mation should reﬂect his/her unconscious reaction to

the task, which should be dependent on his/her per-

sonal experience and moment by moment situation.

The second topic of this section described in Sec-

tion 2.2 is the amount of effort expended in solv-

ing the task. This is again estimated by using bio-

metric information, more speciﬁcally using the eye

movement patterns that should characterize when the

learner conﬁrms that the problem has been solved.

Directly observable behavior such as completion of

entering the answer by using a keyboard could be

used for estimating the above, but the biometric in-

formation should be more accurate than the latter for

estimating it.

2.1 Pupil Diameter: Estimating

Subjective Task Difﬁculty

There are two types of changes concerning the pupil;

mydriasis (dilation of the pupil) and miosis (contrac-

tion of the pupil of the eye) depending on the amount

of light entering the eye and the mental state. It has

been reported that the relationship between the pupil

diameter and the mental state of a person shows a

large mydriasis with the passage of time when a men-

tal stress load is applied, but shows different changes

depending on the type of stress (Taba et al., 1996).

In addition, it has been reported that when the sub-

ject was shown a video in which the object gradually

appeared, mydriasis had already occurred before the

inspiration or new perception was reported (Suzuki

et al., 2018).

These ﬁndings could be used for deriving relation-

ships between the states of efforts and the degree of

task difﬁculties; the amount of mydriasis is propor-

-3

-2.5

-2

-1.5

-1

-0.5

0.5

-1 0 1 2 3 4 5

Δ pupil(arb. unit)

duration time (arb. unit)

baseline B

pattern I

pattern II

pattern III

NPR

, A

NPR

)

max

, A

max

)

Figure 1: The relation between duration time and ∆ pupil

diameter.

tional to the degree of task difﬁculty, i.e., the more

difﬁcult the task becomes, the larger the mydriasis be-

comes due to increased cognitive processing.

Figure 1 schematically shows the relationships be-

tween the changes in pupil diameter and the duration

times. The vertical axis represents the variation of

the pupil diameter denoted as A with reference to the

baseline B

where i denotes the index of the partic-

ipant of the experiment. The horizontal axis repre-

sents the duration time t where t = 0 corresponds to

the time when the stimulus was provided. The points

NPR

and P

max

characterize the pupil reaction to the

input stimuli; P

NPR

corresponds to so-called “near

pupil response,” and P

max

is the point where the an-

swer to the problem is conﬁrmed.

Figure 1 illustrates three patterns that are differen-

tiated by the task solving situations depending on the

values of A

max

relative to B

. The respective partici-

pant’s states would be estimated as follows:

(I) A

max

> B

(top); severely unexpected discovery

of solution with intensive thinking.

(II) A

max

∼ B

(middle); weak unexpected discov-

ery of solution with moderate thinking.

(III) A

max

< B

(bottom); no unexpected discovery

of solution or idle. Gives up to ﬁnd the candi-

date answer.

2.2 Answer Selection Time for

Assessing Subjective Task Efforts

and Strategies

In arithmetic calculation tasks, there are numer-

ous steps to select the answer. For example,

Lebier (Lebiere, 1999) suggested that there are two

basic strategies to conﬁrm the answer: simple retriev-

ing and calculation answer. After choosing strategy,

participants execute the task. The termination condi-

tion of the arithmetic calculation task is divided into

the case where answer conﬁrmation is possible and

not possible. If answer conﬁrmation is possible, the

HUCAPP 2022 - 6th International Conference on Human Computer Interaction Theory and Applications

242

task is completed by selecting either the correct or in-

correct answer. If answer conﬁrmation is not possible,

the task ends due to giving up on the task.

In the experiment referred in this paper, the arith-

metic calculation task is performed within a ﬁxed time

of t

task

seconds in order to induce the partial strategy

selection time to be performed in a short time or to

give up. The difﬁculty level L

of the task presented

to the partial is set to the following three types in light

of the arithmetic teaching guidelines in Japan:

1. Those that can be solved almost easily within t

task

seconds,

2. Those that are thought to be multi-decomposable

in about t

task

seconds, and

3. Those that are almost unlikely to be solved in t

task

seconds.

For retrieving strategy, the process of selecting an-

swers basically matches the pattern of perception in-

formation and memory information. In this sense, the

answer selection time for the arithmetic calculation

task is dependent on whether the participants’ ﬁnish

the pattern matching. If the participant is accustomed

to doing calculations on a regular basis, the pattern

matching will take less time. However, participants

may ﬁnd it uninteresting because it is a dry task. If

not, since it is difﬁcult for participants to perform pat-

tern matching on the task, they will continue matching

for a long time or give up matching. In any case, the

participant needs to change the strategy from retriev-

ing to calculation, in order to select the right answer.

With regard to calculation strategy, there are mul-

tiple steps for answer conﬁrmation. Since the calcu-

lation strategy is a procedural task, it will take more

time for a participant than a simple search, but it will

feel like an intellectual task.

On the other hand, it is also assumed that the

partial cannot answer conﬁrmation by retrieving or

calculation for the arithmetic calculation task. The

existence of inference, which is a third strategy, is

conceivable so as not to give up answer conﬁrma-

tion easily. In other words, when answer conﬁrma-

tion is possible for some digits, answer conﬁrmation

is performed from the candidate presented based on

that information. In the case of answer conﬁrmation

by inference, answer conﬁrmation needs to be per-

formed only for some digits, and thus can be com-

pleted quickly compared to retrieving and calculation.

From the above, the answer selection time T

can be

an index of how the partial selection was performed.

Figure 2 shows the relation between duration time

and eye movement which is represented by AOI. The

participants’ eye movement will be 3 steps:

• watch the given task (Zone A in Figure 2),

Area

Of Interest [arb. Unit]

Duration Time[arb. Unit]

Zone A Zone B Zone C

Figure 2: The relation between duration time and eye move-

ment which represented AOI.

• Work on the task and select the answer (Zone B in

Figure 2),

• stare at the answer (Zone C in Figure 2).

In this paper, T

is deﬁned by the boundary Zones

A and B. When participants are watching the given

task, they concentrate on performing the task; this

means that they cannot afford to look away. In a

sense, participants’ eye movements are for engaging

in the speciﬁc area. At the time of ﬁnishing the given

task, they prepare to select the answer candidate. The

reason why we set T

as the boundary Zones of A and

B is that we save time with regard to participants who

ﬁnished the task and began to select the answer can-

didate. The relationships between the strategy which

participants selected and T

are as follows:

• Pattern D: short time T

, Zone B area, and long

time Zone C. Selected retrieving strategy with

high conﬁdential of answer.

• Pattern E: long time T

, middle time of Zone B,

and short time Zone C. There are two possibili-

ties. (1) selected calculation strategy with high

conﬁdential of answer, and (2) selected inference

strategy.

• Pattern F: short time T

, long time of Zone B, and

no show of Zone C. Gave up to ﬁnd answer.

Through these consideration, we analyze the re-

lation between ∆A and T

, which indicate the partic-

ipants’ condition, and whether they perceive the task

as dry/intellectual issue.

3 SUZUKI ET

AL.’S EXPERIMENT (Suzuki

et al., 2021)

The purpose of this study was to conduct a basic anal-

ysis to estimate the difﬁculty level of a task for learn-

ers by designing calculation tasks with three difﬁculty

levels and measuring biometric data of learners while

Analysis of the Relationship between Subjective Difﬁculty of a Task and the Efforts Put into It using Biometric Information

243

Figure 3: Flow of the experiment.

they were working on the task. This section reviews

the method of the experiment and the main results

concerning the EEG data, which motivated us to con-

duct re-analysis of the pupillary dilation in conjunc-

tion with the times necessary for the participants to

conﬁrm their answers.

3.1 Overview of the Experiment

3.1.1 Equipment

EPOC+ was used for measuring EEG data of the par-

ticipants while engaging in the arithmetic tasks with

the sampling rate of 128 Hz. Tobii Pro Nano was used

for gaze measurement. The sampling rate was 60 Hz.

The monitor used in this experiment was 21.5 inches

with a resolution of 1920 × 1080 pixels. It was set

up so that the distance between the monitor and the

subject was about 57.3 cm. Tobii Pro Lab was used to

conduct the task.

3.1.2 Participants

18 undergraduate and graduate students in their teens

and twenties participated in the experiment. This ex-

periment was approved by the Ergonomic Experiment

Ethics Committee of Nagaoka University of Technol-

ogy.

3.1.3 Procedure

The experimental ﬂow is shown in Figure 3. This ex-

periment was conducted in a cycle in which the gaz-

ing point was displayed at the center of the monitor

for 2,000 msec, followed by the computation task and

answer choices for 5,000 msec.

Three task levels were set for the arithmetic prob-

lems as shown in Table 1 with the following expected

times to solve the problems:

• Normal tasks were expected to be answered in 0.5

sec,

• Easy tasks, less than 0.5 sec, and

• Hard tasks, more than 0.5 sec.

Table 1: Level of calculation task.

Level Contents

Easy Addition of two numbers of three dig-

its without carry-overs

Normal Addition of two numbers of three dig-

its with carry-overs

Hard Multiplication of 3-digit and 2-digit

numbers

The participants were instructed to press a key

once to start the experiment. During the experiment,

they were required to solve arithmetic problems one

by one, and to gaze at one which matched with what

they thought was the answer for the problem, among

the four alternatives.

3.2 Overview of the Main Results

The EEG data for the period of 1500 msec before

and 4500 msec after the presentation of the task

were analyzed by using EEGLAB version 2019.0 and

EEGLAB version 2019.1. The baseline was used for

analysis. Suzuki et al. (Suzuki et al., 2021) speciﬁ-

cally focused on the EEG data known as Fmθ (Yam-

aguchi, 2008) expressed at 67 Hz from the midline of

the frontal lobe. They obtained the frequency spectra

of F3 and F4 channels closest to the midline of the

frontal lobe.

In this experiment, participants responded not by

pressing buttons but by gazing in order to reduce the

effect on the EEG as much as possible. Therefore,

their answer status was analyzed from their gaze data.

The participant calculates the question, then looks for

the number that matches the answer, and stares at the

one with the correct number. The sum of the time

spent gazing at the correct number and the time spent

gazing at the wrong number was calculated.

In the Easy and Normal conditions, the correct and

wrong numbers were clear and the distribution char-

acteristics were similar. The total time spent looking

at the wrong number in one trial was often less than

0.5 seconds, and the task was viewed within 4.5 sec-

onds after presentation. In Hard condition, the distri-

bution of gazing at the wrong number and the correct

number was similar, and it was difﬁcult to determine

the answer chosen by the participant from the gaze

HUCAPP 2022 - 6th International Conference on Human Computer Interaction Theory and Applications

244

data. These results indicate that the correct answer

was given when the gaze was on the correct number

for more than 0.5 sec, a feature observed in the Easy

and Normal. A reading of 0.5 sec or less, but after 4.5

sec, was also considered a correct answer.

4 ANSWER SELECTION AND

PUPILLARY DILATION

4.1 Answer Selection Time

The participant’s task was to mentally calculate the

answer to a given arithmetic problem. The problem

was displayed in the upper center of the screen. The

answer to that question was one of the four choices

given in two rows and two columns at the bottom

of the screen. It was assumed that after seeing the

problem, the participant would complete this task by

deriving an answer using mental arithmetic, ﬁnding

a match among the four options, and staring at it.

Based on this task achievement process, we devised

a method to derive the answer selected by the partici-

pant from the recorded eye movement data.

Figure 4: Angle from the center of the answer numbers.

Among the eye tracking data measured by the eye

tracker, we analyzed those wherein the type of eye

movement was “Fixation.” The gaze data, which were

recorded as two-dimensional data using the coordi-

nates of the monitor (x, y), were converted to one-

dimensional data as shown in Figure 4; a gaze point

was represented as its angle θ from the center of the

four answers denoted as (x

, y

) as follows:

θ = tan

−1

− y

x − x

(−180

◦

≤ θ ≤ 180

◦

) (1)

A participant’s gaze was initially directed to the

question and then on one of the answers s/he chose.

It then directed participants to the area where the

choices for the correct answer were presented. When

the participant’s gaze stayed on the answer area

longer than 100 msec, the entry time was recorded

as the answer selection time.

Figure 5: Difference from the median.

4.2 Pupil Diameter Variation

The pupil diameter data measured by the eye tracker

was analyzed by linear interpolation using the eye

movement type of Fixation. The baseline was deﬁned

as the average pupil diameter during the 500 ms im-

mediately preceding the presentation of the tasks, and

the pupil diameter data were calculated as the varia-

tion from the baseline (Tobii-AB, 2021).

The pupil diameter tended to contract signiﬁcantly

immediately after the task was presented (between

1∼2 sec). This is thought to be a convergence reac-

tion caused by near vision effect. Therefore, the pupil

diameter variation was deﬁned as the range from the

minimum value during the 1∼2 sec, when the conges-

tion reaction was considered to have occurred, to the

maximum value thereafter.

5 RESULT

The answer selection time and the pupil diameter vari-

ation were calculated for tasks of all participants that

were judged to be correct. The median value for

each participant was then calculated, and the differ-

ence between the answer selection time and the pupil

diameter variation for each task and the median value

for each participant was calculated. The results are

shown in Figure 5. Five areas are identiﬁed in Figure

5 as follows:

Area 1. Pupil diameter variation is more than average

and answer selection time is above average.

Area 2. Pupil diameter variation is below average and

answer selection time is later than average.

Area 3. Pupil diameter variation is less than average

and answer selection time is below average.

Area 4. Pupil diameter variation is above average and

answer selection time is earlier than average.

Area 5. Belonging to the ellipse with four points as

vertices, where the pupil diameter variation

is ±0.1 mm, and the answer selection time is

±0.6 sec.

Analysis of the Relationship between Subjective Difﬁculty of a Task and the Efforts Put into It using Biometric Information

245

Figure 6: Quantity for each area.

The results of the count are shown in Figure 6.

The number of trials for each condition estimated to

be correct was 1203 for Easy, 764 for Normal, and

296 for Hard. The average values divided by the num-

ber of areas 5 were 240.6 for Easy, 152.8 for Nor-

mal, and 59.2 for Hard. In the Easy condition, Area

5 was the most common area, and Areas 3 and 4

were more common than average. The Easy condi-

tion exists mostly near the median of the participants.

The difﬁculty level of the problems should have been

higher than that of the Easy condition, so it can be

said that this was the correct attitude to take. In addi-

tion, the change in pupil diameter tended to be larger,

suggesting that the students were concentrating more

on solving the problem than in the other conditions.

In the Hard condition, Area 3 had the most trials and

Area 4 had more trials than the average. Trials judged

to be correct in the Hard condition, moved pupils’

eyes to the answer numbers more quickly than in the

other conditions, suggesting that they were not taking

the task seriously. The degree of change in pupil di-

ameter tended to be less than the average, suggesting

that the participants were not taking the calculation

seriously.

As shown in Figure 7, the results of Suzuki et al.’s

analysis showed high amplitudes in the range of 6 to

7 Hz in the Normal condition, suggesting that Fmθ

could be detected without problems. Pupil diame-

ter variation was also higher in the Normal condition,

suggesting that these two indicators are related.

6 TOWARD APPLICATION TO

EFFECTIVE EDUCATION:

RELATIONSHIP BETWEEN

ARCS MODEL

In this section, we discuss how to apply these results

towards effective education. One of the famous mo-

tivation models is the ARCS model which was pro-

posed by Keller (Keller, 1983; Keller, 1987). Accord-

ing to ARCS model, learner’s motivation is enhanced

by four categories of variables synthesis (Keller,

1987): 1) Attention, 2) Relevance, 3) Conﬁdence, and

4) Satisfaction. Due to its high practicality, the ARCS

model is used in a wide range of ﬁelds such as training

design and teaching material development in compa-

nies, including educational places such as universi-

ties.

Based on the ARCS model, we focus on the cat-

egory of conﬁdence and satisfaction. In procedural

tasks, the learners’ behavior can be represented as be-

low. Before the learner cannot learn the procedure,

when the learner is presented a task, ﬁrstly they try to

candidate the difﬁculty of the task. If the task expects

the task to be easy, the learner will think that the task

can be performed without difﬁculty. If not, the learner

will think that it can be performed with difﬁculty. We

regard this as the process of “candidate”. On the other

hand, the degree of agreement between the difﬁculty

level felt when the task is actually completed and the

expected difﬁculty level is directly related to the satis-

faction level felt by the learner. We regard this as the

process of “assessment”.

Candidates and assessments can be associated

with conﬂicts and satisfactions in the ARCS model.

“Candidates” predict whether the learner will have

the expectation that he or she will be able to do it,

and as a result, the probability that the task will be

successful. The success of a task is equivalent to the

learner’s successful experience. If the learner’s suc-

cessful experience leads to a psychological reward of

”good to do”, it will in turn lead to the learner’s “sat-

isfaction”.

Table 2 represents the criterion of conﬁdence

or satisfaction. The table represents classiﬁca-

tion the participants’ condition according to the

match/mismatch between the L

candidate when the

task is presented and the L

assessment after the task is

executed. Let the candidate L

be L

t,c

and the assessed

be L

t,a

. Pattern A and pattern C are cases where

the L

t,c

and L

t,a

do not match. Pattern A is L

t,c

< L

t,a

though the candidate wrongly estimates the task-ease

assessment. This is an opportunity to deny one’s abil-

ity, so satisfaction is expected to be low. Pattern C

is L

t,c

> L

t,a

, though the candidate wrongly estimated

the ease of the task. This is an opportunity to feel

the improvement of one’s own ability, and as a result,

satisfaction is expected to increase. Pattern B and pat-

tern D are cases where the L

t,c

and L

t,a

match. Pattern

B shows a successful experience with almost no mo-

tivation. Therefore, participants will only attain low

satisfaction for high conﬁdence, which means it is no

surprise that the task is ﬁnished. Pattern D shows the

failure experience that difﬁcult tasks could not be ex-

ecuted after all. Therefore, although the conﬁdence

HUCAPP 2022 - 6th International Conference on Human Computer Interaction Theory and Applications

246

Figure 7: The results of EEG analysis by Suzuki et al. The results of the time-frequency analysis for each condition are

shown, as well as the intervals where a signiﬁcant difference (without multiple comparison correction) was obtained at the

5% level between the conditions. The color of the ﬁgure means that the higher the power compared to the baseline, the redder

the color, and the lower the color, the bluer the color. The horizontal axis of the ﬁgure is time (msec) and the vertical axis is

frequency (Hz).

Table 2: Categorization of candidate/assessment difﬁculty Level and participants’ condition.

assessment

high low

candidate high Pattern D: Pattern C:

feel difﬁcult, give up to solve feel difﬁcult, can solve with ease

low Pattern A: Pattern B:

not motivated, feel difﬁcult to solve not motivated, can solve with ease

is low, there will be no decrease in satisfaction due to

the experience of failure.

One of the similar tasks is the score reading pro-

cess. As shown by Nakahira and Kitajima (Nakahira

and Kitajima, 2017), the reading recognizes and re-

produces the pitch, interval, note value, and motif af-

ter perceiving the representation. Of these, for pitch,

interval, and note value, musical information is ac-

quired while selecting one of the following two strate-

gies. (1) Matching of memorized representation and

perceptual information, (2) Judgment of note type and

note value according to procedure based on partially

memorized reference note.

Participants who are accustomed to score reading,

memorize almost all musical note representations and

use them on a regular basis. Therefore, it is faster to

read the note type and pitch by adopting the strategy

(1). In the case of a complicated score or participants

who are not accustomed to reading notes, it is nec-

essary to adopt strategy (2) to read the pitch, inter-

val, and note value. If the participants adopt strategy

(1), they can understand the score structure in a short

time. From the viewpoint of playing an instrument,

they will think that the score is easy to play. How-

ever, it will often be considered a dry task. When

adopting strategy (2), it takes a long time to under-

stand the structure of the score. But from the view-

point of playing an instrument, it may be difﬁcult to

play. In that case, participants will give up playing the

music or practice playing with the knowledge that it

will take time. In particular, in the case of (2), there

is a possibility that the partial completion can be used

to judge the difﬁculty level in order to provide music

at a level that does not cause participants to quit.

7 CONCLUSION

In this study, we analyzed the relationship between

the difﬁculty of arithmetic using pupil diameter data

and eye movement data and the state of work. We

designed a plane view centered on the answer con-

ﬁrmation time estimated from the eye movement and

∆A. On the plain, we plotted biometric information

to categorize the condition of the participants corre-

sponding to the 5 areas which set the vicinity of the

origin and the four quadrants for the two axes.

As a result, both T

and ∆A were characterized by

the difference in labeled L

. The values of both T

and

∆A increased in the order of Hard, Easy, and Normal

modes. From the scatter plot, we found that the data

for the low difﬁculty tasks were most concentrated in

the Areas 3 and 5, the data for standard difﬁculty tasks

were concentrated in the Areas 1 and 2, and the data

for high difﬁculty tasks were mostly concentrated in

the Areas 3 and 4. These suggested that both T

and

∆A, as biometric information, were suitable quanti-

ties for estimating the states of the participants’ con-

ﬁdence and satisfaction. Based on these results, our

future tasks would be to improve the accuracy of esti-

mation of participants’ conditions for any given tasks.

ACKNOWLEDGEMENTS

This work was partly supported by JSPS KAKENHI

Grant Number 19K12232, 19K12246 and 20H04290.

MH also wants to thank to Nagai N · S Promotion

Foundation For Science of Perception for their ﬁnan-

tial support.

Analysis of the Relationship between Subjective Difﬁculty of a Task and the Efforts Put into It using Biometric Information

247

REFERENCES

Keller, J. M. (1983). Motivational design of instruction.

In Reigeluth, C., editor, Instructional-design theories

and models: An overview of their current status. Hills-

dale, NJ: Lawrence Erlbaum.

Keller, J. M. (1987). Strategies for stimulating the motiva-

tion to learn. Performance + Instruction, 26(8):1–7.

Lebiere, C. (1999). The dynamics of cognition: An ACT-

R model of cognitive arithmetic. Kognitionswis-

senschaft, 8:5–19.

Nakahira, K. and Kitajima, M. (2017). Understanding re-

lationships between reading behavior and difﬁculty

level of musical score based on cognitive-behavioral

science - competency level evaluation via musical

score reading processes. In Proceedings of the 12th

International Joint Conference on Computer Vision,

Imaging and Computer Graphics Theory and Appli-

cations, volume 2, pages 67–74.

Suzuki, R., Akimoto, Y., and Nakahira, K. T. (2021). Dif-

ﬁculty estimation based on combined EEG and Eye

tracker measurements (in Japanese). The 83th Na-

tional Convention of IPSJ, pages 181–182.

Suzuki, Y., Minami, T., and Nakauchi, S. (2018). Associ-

ation between pupil dilation and implicit processing

prior to object recognition via insight. Scientiﬁc Re-

ports, 8.

Taba, N., Takara, T., and Hoshino, K. (1996). Pupillary

diameter and heart rate under active and passive stress

(in japanese). ITE Technical Report, 20(39):75–79.

Tobii-AB (2021). Are pupil size calcu-

lations possible with tobii eye track-

ers? https://www.tobiipro.com/ja/service-

support/learning-center/eye-tracking-essentials/is-

pupil-size-calculations-possible-with-tobii-eye-

trackers (accessed 2021.06.16).

Yamaguchi, K. (2008). Fmθ and alpha 2 rhythms evoked

by mental activity. Research bulletin of the Faculty of

Humanities, 29(1):67–77.

HUCAPP 2022 - 6th International Conference on Human Computer Interaction Theory and Applications

248