You are on page 1of 9

Food Research International 172 (2023) 113119

Contents lists available at ScienceDirect

Food Research International


journal homepage: www.elsevier.com/locate/foodres

Robot versus human barista: Comparison of volatile compounds and


consumers’ acceptance, sensory profile, and emotional response of
brewed coffee
Seyeong Park a, b, Min Kyung Park a, JeongAe Heo a, Ji-sun Hwang a, Sungjae Hwang c,
Daekwang Kim c, Seo-Jin Chung b, Han Sub Kwak a, d, *
a
Food Processing Research Group, Korea Food Research Institute, Wanju-gun 55465, Republic of Korea
b
Department of Nutritional Science and Food Management, Ewha Womans University, Seoul 03760, Republic of Korea
c
XYZ Co., Seoul 04799, Republic of Korea
d
KFRI School, University of Science and Technology, Wanju-gun 55465, Republic of Korea

A R T I C L E I N F O A B S T R A C T

Keywords: The increasing trend of integrating robots into the food industry has sparked debates regarding their potential
Food tech influence on consumer attitudes toward food technology. This study investigated volatile compound profiles via
Food robot gas chromatography–mass spectrometry (GC–MS), consumer acceptability, sensory profiling, and emotional
Coffee
responses of consumers toward coffee samples brewed by robot and human baristas. Moreover, the effect of the
Food technology neophobia
Consumer acceptance
robot experience on food technology neophobia (FTN) was investigated. The principal component analysis of the
Emotions volatile compound profiles revealed that the samples by the robot barista exhibited a higher degree of similarity
compared to those prepared by the human barista. The range of relative standard deviations of volatile com­
pounds from the robot barista brewed coffee was 1.4–83.1% and the variation was smaller than that of the
human barista, which was 5.0–118.3%. Participants had a significant decrease in FTN scores after evaluating the
robot-brewed coffee (p < 0.05), but there was no significant difference in FTN scores before and after evaluating
the coffee brewed by the human barista (p > 0.05). Sensory evaluation studies revealed no significant differences
in acceptability ratings and purchase intentions between the two groups (p > 0.05). However, emotional re­
sponses to the coffee samples significantly varied, with the robot-brewed coffee inducing more dynamic and
positive emotions and the human-brewed coffee inducing more static and positive emotions (p < 0.05). Overall,
this study provides valuable insights into consumer attitudes toward food robot service to humans and indicates
that consumer’s experience with food robots may significantly reduce FTN (p < 0.001).

1. Introduction Korea, cooking robots that fry chicken, cook noodles, and brew coffee
are being introduced into restaurants and cafes (Cho, 2022). The Korean
New trends in the food industry have emerged owing to the outbreak military has deployed cooking robots that can be used for frying, stir-
of COVID-19. The most notable changes are increases in consumption of frying, and cooking soup and rice to overcome decline in manpower
meal kits and home meal replacements, online orders, take-outs, and (Choi, 2022).
drive-through orders in vehicles to avoid infection arising from human Compared to humans, food robots can be attractive, smartly identify
contact (Lee and Ham, 2021). In particular, COVID-19 has accelerated required refreshment orders, and work 24/7 without a payroll, thereby
the development of food technology related to robotics and automation. reducing additional manpower. Moreover, when juxtaposed with
Robots are being applied in food production, cooking, and food services. human counterparts, food robots excel in executing tasks with enhanced
Robotic food services are already applied in limited areas, such as speed, unwavering consistency, and unparalleled accuracy (Abdelhakim
delivering food to buildings (Park, 2019) and airports (Hwang, 2022). et al., 2023; Asif et al., 2015; Ivanov and Webster, 2017). In the coffee
Robots have also infiltrated the realm of cooking. In the Republic of industry in particular, there is a growing interest in the adoption of

* Corresponding author at: Food Processing Research Group, Korea Food Research Institute, Wanju-gun 55465, Republic of Korea.
E-mail address: hskwak@kfri.re.kr (H.S. Kwak).

https://doi.org/10.1016/j.foodres.2023.113119
Received 2 March 2023; Received in revised form 7 June 2023; Accepted 9 June 2023
Available online 12 June 2023
0963-9969/© 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
S. Park et al. Food Research International 172 (2023) 113119

Fig. 1. Coffee preparation by a robot barista (a) and sensory evaluation (b), and coffee preparation by a human barista (c) and sensory evaluation (d).

robotics and automation, centered around robotic baristas (Sung and measure consumer acceptance of new technologies applied for food (Cox
Jeon, 2020). Robotic baristas improve the accuracy, consistency, and and Evans, 2008). Coppola et al. (Coppola et al., 2014) suggested that
efficiency of coffee making, creating new experiences and value prop­ the FTNS may be a better tool than the food neophobia scale (FNS) for
ositions for consumers (Kim et al., 2021a). Robots masterfully brew predicting consumers’ willingness to try new food technologies because
coffee using identical brewing procedures, guaranteeing a consistently FTNS focuses on technology rather than food. Initially, FTNS was used to
exceptional sensory profile across numerous servings. This consistency study neophobia regarding new technologies applied in food production
in coffee quality is rather crucial in commercial settings for customer and nutritional fortification. Several studies have focused on food pro­
satisfaction and brand reputation. Furthermore, robot brewing enables cessing technologies, such as pasteurization, modified atmosphere
the provision of top-tier coffee round the clock, catering to the needs of packaging, high-pressure processing, genetic modification, and nano­
customers in cafés and highway resting areas without compromising on technology of food (Chen et al., 2013; Cox and Evans, 2008; Evans et al.,
quality (2022). 2010; Matin et al., 2012), and nutritional aspects, such as reduced fat
Gradually, some restaurants are moving from human–human inter­ and functional and organic food (Caracciolo et al., 2011; Coppola et al.,
action to human–robot interaction. Recently, studies have been con­ 2014; Verneau et al., 2014). Recently, consumer studies on three
ducted on robots that act as chefs in restaurants (Fusté-Forné, 2021; dimensional (3D) printed food (Caulier et al., 2020; Feng et al., 2022;
Seyitoğlu et al., 2021; Xiao and Zhao, 2022; Zhu and Chang, 2020). (Lee et al., 2021) and in vitro meat (Baum et al., 2021; (Heidmeier and
Fusté-Forné (Fusté-Forné, 2021) and Seyitoğlu et al. (Seyitoğlu et al., Teuber, 2022; (Krings et al., 2022), which have the potential to solve
2021) reported ambivalent attitudes towards robot chefs through an global issues by stimulating sustainability transition, have been con­
interview-based survey to assess consumers’ perceptions about robot ducted using FTNS. Although several consumer studies have been con­
chefs in restaurants. In their studies, participants showed positive atti­ ducted on the application of various food technologies, studies on
tude to robot chefs because of consistent and fast cooking and cleanli­ consumers based on direct experience of food robots are lacking. In
ness. In contrast, negative attitudes were noticed regarding a lack of particular, to our best knowledge, studies in which consumers actually
cooking details and interaction with consumers. Xiao and Zhao (2022) observe the process of making food by a cooking robot and subsequently
investigated the factors influencing consumers’ quality predictions for taste the prepared food have not been conducted.
robot-cooked foods. They found that despite participants expressing a Therefore, this study aimed to evaluate the influence of real-life
positive perception of the food quality prepared by robot chefs, there experience with a robot barista on consumers’ attitude towards food
remained a prevailing attitude that robot-made foods were of inferior technology. Specifically, this study compared volatile compound pro­
quality compared to those prepared by human chefs, irrespective of the files using gas chromatography–mass spectrometry (GC–MS), consumer
complexity of the cooking process. Several studies have been conducted acceptability, sensory profiling, and emotional responses of consumers
on consumers’ attitude towards cooking by robots, using images such as towards coffee samples brewed by robot and human baristas. In addi­
screenshots from cooking videos featuring human chefs or robot chefs tion, this study investigated how robot experience of brewing coffee
(Xiao and Zhao, 2022)or video clips compiled from YouTube videos influences changes in FTN.
showcasing robot in restaurants (Seyitoğlu et al., 2021). However, there
has been a lack of direct assessment regarding consumer acceptability
and attitudes towards cooking performed by robots.
The food technology neophobia scale (FTNS) was designed to

2
S. Park et al. Food Research International 172 (2023) 113119

2. Material and method Table 1


Demographic profiles of participants.
2.1. Analysis of volatile compounds Category RBC1) HBC
(n = 67) (n = 75)
2.1.1. Chemicals and coffee brewing Gender
Methanol and 2-methylheptan-3-one were purchased from Sigma- Male 23 (34.3%) 24 (32.0%)
Aldrich (St. Louis, MO, USA). Coffee brewing was conducted at a local Female 44 (65.7%) 51 (68.0%)
café, where a robot barista was installed (LOUNGE’X, Daejeon, Republic Age range (years)
19–29 36 (53.7%) 48 (64.0%)
of Korea). Coffea arabica (Ethiopia Koke Honey: natural processing and 30–39 18 (26.9%) 15 (20.0%)
1,850–2,150 m of elevation) was used. The green beans were roasted by 40–49 10 (14.9%) 9 (12.0%)
a third-party professional supplier (A Day at the Coffee Farm, Seoul, 50–59 3 (4.5%) 3 (4.0%)
Republic of Korea). This coffee is regularly brewed at the café. Samples Occupation
Student 21 (31.3%) 31 (41.3%)
were prepared using a robot barista (XYZ Co., Seoul, Republic of Korea)
Housewife 12 (17.9%) 8 (10.7%)
and by a professional barista using the hand drip method (Fig. 1). In Self-employment 3 (4.5%) 4 (5.3%)
robot barista brewing, programmed brewing procedure created by the Office worker and researcher 7 (10.5%) 13 (17.3%)
manufacturer (XYZ Co.) was used. The process began by heating filtered Sales and service 9 (13.4%) 11 (14.7%)
water (using an HF CSR TWIN A-10/SR-X system by Pentair plc., Lon­ Production worker 1 (1.5%) 0 (0.0%)
Government employee 8 (11.9%) 5 (6.7%)
don, UK) to a temperature of 96 ◦ C. The heated water was then carefully Unemployed 6 (9.0%) 3 (4.0%)
poured in a swirling motion over 30 g of ground coffee. This swirling Educational level
motion was repeated four times, following a spiral pattern from the High school dropout or below 1 (1.5%) 0 (0.0%)
inside to the outside, with respective water quantities of 45 g, 90 g, 88 g, High school graduation 17 (25.4%) 22 (29.3%)
Junior college graduation 4 (6.0%) 10 (13.3%)
and 65 g. The entire process lasted for a duration of 3 min. The human
University graduation 44 (65.7%) 40 (53.3%)
barista brewed coffee according to the brewing method programmed Graduate school graduation 1 (1.5%) 3 (4.0%)
into the robot barista. While brewing the coffee, the human barista Monthly household income1)
carefully regulated the amount of water for dripping by utilizing a scale Less than $2,317.9 31 (46.3%) 32 (42.7%)
for precise measurement. The brewed coffee samples, weighing $2,317.9-$3,863.2 21 (31.3%) 24 (32.0%)
$3,863.2-$5,408.4 9 (13.4%) 6 (8.0%)
approximately 3 g, were subjected to oven drying at a temperature of
>5,408.4 6 (9.0%) 13 (17.3%)
105℃ for a duration of 24 h. Total dissolved solids (TDS) were deter­
2)
mined using the following equation: The dollar shown is based on the exchange rate on December 14, 2022 ($1 =
TDS (%) = 100 – [(W1 – W2) / W1 × 100]. 1,297.5 won).
1)
RBC: Robot brewed coffee, HBC: Human brewed coffee.
Where W1 = weight of the sample before oven drying and W2 =
weight of the sample after oven drying.
derived from the GC column) and/or with a signal-to-noise ratio > 5
2.1.2. GC–MS analysis of volatile compounds in coffee samples were included in the final dataset for multivariate statistical analysis.
A total of six replicates of coffee brewing were conducted using both Relative concentration was obtained by comparing each peak area to
robot and human baristas, with the aim of analyzing volatile compounds that of the internal standard compound. Principal component analysis
present in the brewed coffee. Solid-phase microextraction was used to (PCA) was conducted to discriminate between two different coffees
extract volatile compounds from coffee samples. Coffee sample (3.0 g) based on the profiles of volatile compounds, using SIMCA 16 (Umetrics,
and the internal standard (2 μL, 2-methylheptan-3-one, 100 mg/L in Umea, Sweden). All experiments were conducted six times.
methanol) were placed into a 20-mL headspace amber vial. The vial was
kept at 40℃ for 30 min to obtain an equilibrium state. Divinylbenzene/ 2.2. Sensory evaluation
carboxen/polydimethylsiloxane/fiber (DVB/CAR/PDMS, 50/30 μm,
Supelco, Bellefonte, PA, USA) was inserted into the headspace of the vial 2.2.1. Participants
for 30 min at 40℃, and volatile compounds were thermally desorbed in Participants were recruited through an online trading application of
the injector port of GC–MS for 5 min at 230 ◦ C. local goods (Dangeun Market, Inc., Seoul, Republic of Korea). The
Gas chromatography (7890A, Agilent Technologies, Santa Clara, CA, recruitment information was restricted to the users in Daejeon, Republic
USA) with a mass spectrometer (5975C, Agilent Technologies) was used of Korea. Only coffee drinkers without issues in completing the ques­
to analyze the volatile compounds in coffee. A fused silica capillary tionnaire using a smartphone or computer were eligible to participate in
column (DB-UIWAX, 30 m length × 0.25 mm i.d. × 0.25 μm film this study. Participants’ intentions to participate in the sensory evalua­
thickness, J&W Scientific, Folsom, CA, USA) was used to separate the tion were confirmed using an online survey tool (Naver, Co., Seongnam-
analytes. The flow rate of helium (carrier gas) was 0.8 mL/min. Mass si, Republic of Korea). Respondents, who agreed to provide their per­
spectra were collected in the EI mode at 70 eV, and the mass scan range sonal information and a bank account number for monetary compen­
was m/z 35–350. The oven temperature was maintained at 40 ◦ C for 5 sation, were selected as participants. Participants were asked to fill out
min and then increased to 220℃ at a rate of 4℃/min. Inlet and transfer an online consent form using an electronic contract system (Modusign
line temperatures were 230℃ and 250℃, respectively. Inc., Seoul, Republic of Korea). Among the total participants who signed
the online consent form, 142 out of 152 individuals (93.42%) met the
2.1.3. Non-target analysis of profiles of the volatile compounds eligibility criteria and were able to take part in the sensory evaluation.
For non-target analysis, the data obtained from GC–MS were pre- Demographic information of the participants is presented in Table 1. The
processed using the MS-DIAL platform (ver. 4.92) for deconvolution experiment was approved by the Institutional Review Board of the Korea
and alignment of peaks from the samples. The following parameters Food Research Institute (KFRI).
were applied for MS-DIAL: mass range was between 35 and 350 Da; The
minimum peak height for peak detection was 2,000 amplitudes; the 2.2.2. Sample preparation and sensory evaluation procedure
retention time tolerance for alignment was 0.05 min; and the identifi­ One day prior to the sensory evaluation, participants were asked to
cation score cut-off was 60%. The peak areas were calculated using MS- fill in questionnaires for FTN (Cox and Evans, 2008)and provide de­
DIAL and normalized according to the total ion chromatogram. A total of mographic information using their smartphones, tablet PCs, or laptops.
245 features without the characteristic m/z fragment 73.1 (siloxanes Participants completed a 13-item FTNS using a 7-point scale ranging

3
S. Park et al. Food Research International 172 (2023) 113119

Fig. 2. PCA score plot based on volatile compounds profiles derived from human brewed coffee (HBC, red dots) and robot brewed coffee (RBC, yellow dots).

from 1 (strongly disagree) to 7 (strongly agree). The FTNS assesses dark chocolate, burnt, citrus fruit, nutty, and grain; mouthfeel: thick­
consumers’ attitudes towards foods produced using novel technologies. ness, longevity, astringent, and mouth drying) that were selected based
The scale includes questions related to the necessity of new food tech­ on previous studies (Cotter et al., 2021; Seo et al., 2009; World Coffee
nology (questions 1–6), recognition of risk (questions 7–10), healthy Research, 2017) and reviewed by six sensory scientists who tasted the
choice (questions 11 and 12), and information and media (question 13). samples and deliberated on the appropriate terms to use. Emotional
Total scores on the 13-items FTNS range from 13 to 91, with higher responses of the participants from the beginning of brewing to the end of
scores indicating a greater neophobia or aversion to new food technol­ sensory evaluation were measured. Twenty-five emotional terms in the
ogy (Cox and Evans, 2008). The Korean-translated version of FTNS was EsSense25 list (Nestrud et al., 2016)were assessed using the RATA
not available; therefore, it was directly translated by a sensory scientist method: active, bored, disgusted, energetic, free, merry, peaceful, joyful,
at the KFRI. Five sensory scientists reviewed and revised the Korean pleased, satisfied, wild, worried, good, pleasant, guilty, warm, enthusi­
version of FTNS. The results were collected using Compusense Cloud astic, adventurous, loving, nostalgic, secure, good-natured, calm,
software (Compusense Inc., Guelph, Canada). aggressive, and understanding. Since ’joyful’ and ’pleasant’ have the
The participants were randomly assigned to taste the robot- (n = 67) same meaning when translated into Korean, a total of 24 terms except
or human- (n = 75) brewed coffee samples. They examined the entire ‘joyful’ were used in this study. While using the RATA method for sen­
brewing procedure from the beginning to serving. Coffee samples for sory attributes and emotions, participants were asked to check the terms
sensory evaluation were prepared in the same way as that for volatile they considered appropriate for describing samples and then to rate the
compound analysis. The sample (approximately 80 mL) was poured into intensity of applicable terms using a 4-point structured scale (0, none, 1,
a white paper cup (10 oz) for evaluation. The serving temperature of weak; 2, medium; and 3, strong). After the evaluation of samples, the
coffee was approximately 68–70 ◦ C. participants were once again presented with the FTN questions for
During sensory evaluation at the test location, the participants were further assessment.
asked to fill out a paper ballot for the questionnaire.
The testing was carried out within the café setting, where partici­ 2.2.3. Data analyses
pants had the unique opportunity to witness the coffee brewing process Demographic profiles of the participants were expressed as fre­
performed by either a human or a robotic barista. This scenario is quencies and percentages using descriptive statistics. Differences in the
visually represented in Fig. 1. Each session was assessed by three par­ pre-post FTNS survey within each variable were analyzed using a paired
ticipants. The questionnaire consisted of queries regarding product t-test (p < 0.05). Consumers’ acceptability, purchase intent, sensory
acceptability, sensory profiling, and emotional responses for the coffee attributes, and emotion were analyzed using independent sample t-test
sample. Acceptability for overall liking and likings of color, aroma/odor, (p < 0.05) to determine the differences among samples. In addition, for
taste/flavor, and mouthfeel was assessed using a 9-point hedonic scale visual representation, a bar graph was used for the distribution of
(1, extremely dislike; 5, neither like nor dislike; and 9, extremely like). acceptability and purchase intent for the sample, and a radar chart was
Purchase intent was measured using a 5-point scale (1, I would definitely used for the distribution of emotional responses for each sample. Data
not buy and 5, I would definitely buy). Consumers were then provided were analyzed using XLSTAT v.2021 (Addinsoft, Paris, France).
with a rate-all-that apply (RATA) list consisting of 16 sensory attributes
(aroma: sour, citrus fruit, and burnt; taste/flavor: bitter, sour, sweet,

4
S. Park et al. Food Research International 172 (2023) 113119

compound profiles of the two groups were not significantly different.


To compare the repeatability (n = 6) of each group, the relative
standard deviation (RSD) of each relative concentration was calculated
and expressed as a scatter plot (Fig. 3). The data used to calculate the
RSD percentage were the mean values of six coffee samples prepared
within a day. The RSD of peak areas ranged between 1.4 and 83.1% and
5.0–118.3% in RBC and HBC, respectively. Based on these results, the
distribution of RSD values was larger in the human brewed coffee
samples than in the robot brewed coffee samples. Higher variations in
volatile compounds during repeated brewing were observed in HBC than
in RBC. This indicates that the robot barista was able to brew coffee
more consistently than the human barista.

3.2. Sensory evaluation of coffee

3.2.1. Demographic characteristics


Demographic information of the participants is shown in Table 1. In
Fig. 3. Distribution of repeatability (relative standard deviation, RSD) based on the RBC and HBC groups, the majority were female (65.7 and 68%,
volatile compounds in two different coffee samples obtained from GC–MS. RBC respectively), aged 19–39 years (80.6 and 84%, respectively), and col­
and HBC meant robot brewed coffee and human brewed coffee, respectively. lege students (31.3 and 41.3%, respectively). University graduates (65.7
and 53.3%) were most common. Most households (77.6 and 74.7%) had
3. Results and discussion a monthly income <$3,863.2. In general, the participants in the evalu­
ation of RBC and HBC were similar.
3.1. Comparison of volatile compounds in two different coffee samples
3.2.2. Food technology neophobia
Aroma is mainly related to volatile chemical compounds and is the The paired t-test results of the FTN survey conducted before and after
most important factor in evaluating the quality of coffee. In this study, sensory evaluation are presented in Table 2. In the RBC group, the mean
the profiles of volatile compounds derived from coffee samples prepared values of total scores pre- and post-survey were 42.81 ± 9.87 and 38.85
by robot and human baristas were compared by performing non-target ± 9.55, respectively, showing a significant difference (p < 0.001). The
and multivariate statistical analyses. A target analysis relies on the mean values of total scores pre- and post-survey were 41.09 ± 7.65 and
selected individual analytes only, whereas a non-target analysis can 40.44 ± 9.31, respectively, showing no significant difference (p > 0.05)
provide many possible global fingerprints and comprehensive informa­ in the HBC group. Therefore, the consumers’ experience with robot
tion (Cavanna et al., 2018). This methodology has been applied to verify barista lowered the FTN score. Additionally, the FTN scores in the pre-
the authenticity of various foods such as meat (Barbieri et al., 2020)and survey were relatively lower than those in other countries, such as
olive oil (Kim et al., 2020). Fig. 2 shows the classification of robot Australia (54.4; (Evans et al., 2010), Italy (60.9; (Verneau et al., 2014),
brewed coffee (RBC) and human brewed coffee (HBC) using PCA. RBC and Brazil (47.0; (Vidigal et al., 2015). In these studies, FTNS questions
and HBC were not clearly divided into two groups, and the PCA model were asked prior to the evaluation of new food technologies such as
showed poor performance (R2X = 0.854, Q2 = 0.372) in discriminating pasteurization, modified atmosphere packaging, high-pressure process­
the samples. This can be explained by the fact that the volatile ing, genetic modification, and nanotechnology. The reason for low FTN
scores can be attributed to a low age and large number of university

Table 2
The rating of food technology neophobia scale from the pre- and post-survey.
Item RBC (n = 67) HBC (n = 75)

Pre-survey Post-survey P- Pre-survey Post-survey P-


value2) value
1 There are plenty of tasty foods around so we don’t need to use new food technologies to 2.69 ± 2.04 ± 1.35 ** 2.08 ± 1.11 2.19 ± 1.34 n.s.
produce more. 1.871)

2 The benefits of new food technologies are often grossly overstated. 4.13 ± 1.31 3.90 ± 1.53 n.s. 4.36 ± 1.32 4.24 ± 1.48 n.s.
3 New food technologies decrease the natural quality of food. 2.94 ± 1.46 2.30 ± 1.23 ** 2.71 ± 1.19 2.45 ± 1.29 n.s.
4 There is no sense trying out high-tech food products because the ones I eat are already 2.42 ± 1.35 2.06 ± 1.09 * 2.25 ± 1.16 1.97 ± 0.99 *
good enough.
5 New foods are not healthier than traditional foods. 2.87 ± 1.46 2.45 ± 1.46 n.s. 2.55 ± 1.34 2.51 ± 1.23 n.s.
6 New food technologies are something I am uncertain about. 3.34 ± 1.47 3.19 ± 1.66 n.s. 3.44 ± 1.40 3.44 ± 1.62 n.s.
7 Society should not depend heavily on technologies to solve its food problems. 3.10 ± 1.66 3.16 ± 1.79 n.s. 2.97 ± 1.57 3.49 ± 1.72 **
8 New food technologies may have long term negative environmental effects. 3.27 ± 1.59 3.01 ± 1.50 n.s. 3.28 ± 1.47 3.27 ± 1.65 n.s.
9 It can be risky to switch to new food technologies too quickly. 4.42 ± 1.56 3.73 ± 1.78 ** 4.44 ± 1.45 4.40 ± 1.63 n.s.
10 New food technologies are unlikely to have long term negative health effects (R)3). 4.16 ± 1.42 3.96 ± 1.61 n.s. 3.91 ± 1.31 3.91 ± 1.43 n.s.
11 New products produced using new food technologies can help people have a balanced 3.15 ± 1.42 2.60 ± 1.30 ** 2.91 ± 1.07 2.59 ± 1.31 n.s.
diet (R).
12 New food technologies give people more control over their food choices (R). 2.37 ± 1.15 2.37 ± 1.42 n.s. 2.11 ± 0.94 1.91 ± 0.93 n.s.
13 The media usually provides a balanced and unbiased view of new food technologies (R). 3.94 ± 1.58 4.07 ± 1.80 n.s. 4.09 ± 1.42 4.08 ± 1.62 n.s.
Total 42.81 ± 38.85 ± *** 41.09 ± 40.44 ± n.s.
9.87 9.55 7.65 9.31
1)
Mean ± SD. This items were asked using 7-point Likert scale (1 = strongly disagree, 4 = neither agree nor disagree, and 7 = strongly agree).
2)
P-value correspond to paired t-test to related pre- and post-survey. Significance levels are as follows: (***) P < 0.001; (**) P < 0.01; (*) P < 0.05; n.s. meant non-
significant.
3)
(R) indicates reverse scored items.

5
S. Park et al. Food Research International 172 (2023) 113119

RBC HBC

n.s. n.s. n.s.


9
n.s.2) n.s.
8

6
Ratings

5
n.s.
4

1
Overall1) Color Aroma/odor Taste/flavor Mouthfeel Purchase intent
Acceptability and purchase intent
Fig. 4. Acceptability and purchase intent ratings of the coffee samples. RBC and HBC meant robot brewed coffee and human brewed coffee, respectively. 1)
Acceptability was asked using a 9-point hedonic scale, and purchase intent was measured by using a 5-point scale. 2) Significance levels are as follows: n.s. meant
non-significant.

students as participants in this study. ratings, all acceptance attributes were above five out of nine; therefore,
In the RBC group, a significant difference in the five questions was the participants generally accepted the tasting sample. Even though the
noticed before and after surveys for FTN (p < 0.05). The post-score was participants watched the entire procedure of coffee brewing by a robot
significantly lower than the pre-score (p < 0.05), confirming that FTN or a human, the brewing procedure did not affect acceptance ratings and
decreased after the experience of robot coffee brewing. Particularly, low purchase intent. The difference in the samples of this study was the
ratings in the post-score were associated with the questions related to coffee brewer rather than difference in ingredients or cooking proced­
‘necessary of new food technology’ such as questions 1, 3, and 4 on ure. Brewing by a robot or human did not lead to significant difference in
Table 2. Additionally, low scores were observed for question 9, which the intrinsic quality of samples between RBC and HBC. Since no differ­
was related to recognition of risk, and for number 11, which was related ence in aroma and taste was noticed between the samples, purchase
to healthy choice after robotic coffee brewing experience. In the HBC intention should not vary. Our findings are in line with those of previous
group, two questions showed significantly different results before and studies that evaluated food products prepared using food technology.
after surveys for FTN (p < 0.05). In particular, a higher rating in the post- Kuang et al. (Kuang et al., 2020) investigated the effects of awareness of
survey than that of the pre-survey was observed for question 7 (society food products prepared by applying nanotechnology. In their study,
should not heavily depend on technologies to solve food issues), which participants showed no significant difference in acceptance ratings be­
was not directly related to new food technology. Following the experi­ tween samples, although nanotechnology labels were applied to the food
ence of the human barista brewing, the degree of FTN sentiment products, and 75–86% of the participants were willing to purchase food
increased. This indicates that participants developed a more negative produced with the help of nanotechnology in the future. Feng et al.
attitude towards food technology after witnessing the coffee preparation (Feng et al., 2022) studied the effect of labeling and product information
conducted by a human barista. Previous studies highlighted the on milk chocolate, gummy vitamin supplement form, and baked potato
importance of providing appropriate information for food technology to made using 3D printing technology on consumers’ perception. The
increase consumer acceptance of products made using food technology participants received a brief description of 3D printing technology and
(Laureati et al., 2016; Villegas et al., 2008). In addition, Cattaneo et al. evaluated the samples labeled as conventional and 3D printed. The 3D
(Cattaneo et al., 2019) revealed that FTN was lower when consumers printing technology and food labeling did not induce significant differ­
were provided with information on food technology than when infor­ ences in acceptance. Therefore, if food quality is very similar regardless
mation was not provided. We found that consumers’ exposure to a of whether food technologies are applied, consumers do not seem to
cooking robot significantly reduced FTN (p < 0.05). This indicates that a have discrepancies in acceptance and purchase intent. Additionally,
first-hand exposure to a cooking robot can effectively mitigate con­ Xiao and Zhao. (Xiao and Zhao, 2022) suggested that cooking difficulty
sumers’ hesitancy towards the notion of a robotic cook. should be considered when predicting the quality of human-cooked and
robot-made food. Consumers predict that robot-made food will taste less
3.2.3. Consumers’ acceptance and purchase intent delicious than human-made food in luxury restaurants (Nozawa et al.,
The TDS of RBC and HBC were 2.07 ± 0.07% and 2.18 ± 0.09%, 2022). On the other hand, consumers have liked cooking by robots in
respectively (p < 0.05; results not shown). RBC showed more consistent fast food restaurants, where cooking was relatively simple (Seyitoğlu
brewing quality than HBC. However, this difference did not generate and Ivanov, 2022). In this study, brewing coffee constituted a relatively
any change in acceptance ratings and purchase intent. No significant simple procedure compared to other cooking methods; therefore, the
difference was presented between RBD and HBC samples in all accep­ participants seemed to have less resistance to coffee brewing using a
tance ratings and purchase intent (p > 0.05; Fig. 4). In the acceptance robot. In a previous survey for robot-operated coffee experience, it was

6
S. Park et al. Food Research International 172 (2023) 113119

Table 3 a study by Schouteten et al. (Schouteten et al., 2019), same food items
The intensities of sensory attributes for the coffee samples using RATA method. with different labeling (organic or conventional) were differently eval­
Sensory attributes RBC HBC p-value uated for some attributes depending on the labeling, such as fruit aroma
(n = 67) (n = 75) and flavor in yogurt samples and orange aroma and flavor in orange
O_sour1) 1.90 ± 1.142) 1.63 ± 1.10 n.s.3) juice. Sensory attributes were recognized differently depending on the
O_citrus fuit 0.60 ± 0.89 0.43 ± 0.84 n.s. context of test location (Boutrolle et al., 2007; Niimi et al., 2022;
O_burnt 1.03 ± 0.97 0.79 ± 0.93 n.s. Schouteten et al., 2019). Therefore, evaluation results of sensory prop­
T_bitter 1.19 ± 1.02 1.13 ± 1.06 n.s. erties of samples could be different owing to extrinsic viewpoints rather
T_sour 2.18 ± 0.98 1.80 ± 1.15 *
T_sweet 0.18 ± 0.49 0.23 ± 0.56 n.s.
than intrinsic characteristics of samples themselves. Considering the
F_roasted 0.75 ± 0.96 0.97 ± 1.08 n.s. similarity of attributes between RBC and HBC, the effect of sample
F_dark chocolate 0.69 ± 0.96 0.40 ± 0.74 * preparation (robot vs. human) did not affect the sensory profile of the
F_burnt 1.01 ± 1.07 0.87 ± 0.98 n.s. sample.
F_citrus fruit 0.52 ± 0.84 0.44 ± 0.89 n.s.
F_nutty 0.52 ± 0.88 0.65 ± 0.94 n.s.
F_grain 0.30 ± 0.67 0.24 ± 0.57 n.s. 3.2.5. Emotional responses of samples
M_thickness 1.18 ± 1.09 1.59 ± 1.05 * Among the list of 24 emotional terms based on EsSense25® (Nestrud
M_longevity 2.00 ± 1.02 1.67 ± 1.14 n.s. et al., 2016), emotions from the start of brewing to after drinking coffee
M_astringent 1.00 ± 1.00 0.91 ± 0.98 n.s. were evaluated using the RATA method. Fourteen out of 24 emotional
M_mouth drying 0.49 ± 0.88 0.32 ± 0.66 n.s.
terms were significantly different (p < 0.05) between RBC and HBC
1)
O: odor, T: taste, F: flavor, M: mouthfeel. (Fig. 5). In RBC, the RATA scores of six emotional terms (adventurous,
2)
Mean ± SD. RATA method was asked using 4-point structured scale (0 = worried, interested, pleasant, active, and aggressive) were significantly
none, 1 = weak, 2 = medium, and 3 = strong). higher than those in HBC (p < 0.05). In HBC, the RATA scores of eight
3)
Significance levels are as follows: (*) P < 0.05; n.s. meant non-significant.
emotional terms (loving, nostalgic, warm, good natured, calm, tame,
good, and tender) were significantly higher than those in RBC (p <
found that consumers who had the opportunity to experience robot 0.05). With the exception of worried, which is a negative emotion, most
barista-made coffee expressed surprise, delight, and high levels of of the differences between RBC and HBC were related to positive
satisfaction with the quality of their drinks (Kim et al., 2021b). emotional terms. The emotions in RBC were dynamic and positive,
whereas HBC showed static and positive emotions associated with
3.2.4. Comparison of sensory profiles using RATA relaxation. Consumers’ emotional attitude towards robot services is
The sensory profiles of coffee samples using RATA are shown in changing upon exposure to COVID-19. Prior to COVID-19 (Chan and
Table 3. In general, the participants could not distinguish the differences Tung, 2019; Kattara and El-Said, 2013; Stock and Merkle, 2018), human
in sensory characteristics between samples (p > 0.05). Three out of 16 services were preferred over robot services in hotels. However, since
attributes (sourness, dark chocolate flavor, and longevity mouthfeel) COVID-19, as the acceptance of robots that provide non-contact services
showed significant differences between samples (p < 0.05). According to and help in reducing anxiety of infection through human interaction has

Fig. 5. Radar plot of emotional terms for the coffee samples. RBC and HBC meant robot brewed coffee and human brewed coffee, respectively. RATA method was
asked using 4-point structured scale (0 = none, 1 = weak, 2 = medium, and 3 = strong). Significance levels are as follows: (***) p < 0.001; (**) p < 0.01; (*) p < 0.05.

7
S. Park et al. Food Research International 172 (2023) 113119

increased, some positive feelings have emerged for robot services than References
for human services (Kim et al., 2021a). Additionally, according to Labbe
et al. (Labbe et al., 2015), emotions elicited by the experience of coffee Abdelhakim, A. S., Abou-Shouk, M., Ab Rahman, N. A. F. W., & Farooq, A. (2023). The
fast-food employees’ usage intention of robots: A cross-cultural study. Tourism
brewing can vary depending on the individual’s motivation of con­ Management Perspectives, 45, Article 101049.
sumption. In our study, differences were observed among the positive Asif, M., Sabeel, M., Mujeeb-ur Rahman, K. (2015). Waiter robot-solution to restaurant
emotional terms because the test location was a relatively high-end café, automation 2015 At Wah, Pakistan 14 15.
Barbieri, S., Cevoli, C., Bendini, A., Quintanilla-Casas, B., García-González, D. L., &
and the attitude of the participant towards coffee drinking was closer to Gallina Toschi, T. (2020). Flash gas chromatography in tandem with chemometrics:
hedonic than to utilitarian. We found that the interaction with the robot A rapid screening tool for quality grades of virgin olive oils. Foods, 9(7), 862.
barista was highly enjoyable and intriguing for participants. Addition­ Baum, C. M., Bröring, S., & Lagerkvist, C.-J. (2021). Information, attitudes, and consumer
evaluations of cultivated meat. Food Quality and Preference, 92, Article 104226.
ally, we observed that the element of playfulness in the robot consid­ Boutrolle, I., Delarue, J., Arranz, D., Rogeaux, M., & Köster, E. P. (2007). Central location
erably influenced their overall satisfaction (Kim et al., 2021b; Sung and test vs. home use test: Contrasting results depending on product type. Food Quality
Jeon, 2020). and Preference, 18(3), 490–499.
Caracciolo, F., Coppola, A., & Verneau, F. (2011). Validation of a psychometric scale to
measure consumers’ fears of modern food technologies. Proceedings in Food System
4. Conclusions Dynamics, 160–174.
Cattaneo, C., Lavelli, V., Proserpio, C., Laureati, M., & Pagliarini, E. (2019). Consumers’
attitude towards food by-products: The influence of food technology neophobia,
This study offers valuable insights into consumer attitudes toward education and information. International Journal of Food Science & Technology, 54(3),
food robots for cooking. The results indicate that direct experience with 679–687.
a food-making robot can effectively reduce consumers’ reluctance to­ Caulier, S., Doets, E., & Noort, M. (2020). An exploratory consumer study of 3D printed
food perception in a real-life military setting. Food Quality and Preference, 86, Article
ward the foods, as demonstrated by the decrease in FTN scores. The 104001.
consistency of volatile compound profiles in robot barista brewed coffee Cavanna, D., Righetti, L., Elliott, C., & Suman, M. (2018). The scientific challenges in
suggests that robot baristas have the potential to be a reliable and effi­ moving from targeted to non-targeted mass spectrometric methods for food fraud
analysis: A proposed validation workflow to bring about a harmonized approach.
cient alternative to human baristas. However, the emotional responses
Trends in Food Science & Technology, 80, 223–241.
to the coffee samples emphasize the need to consider not only the Chan, A. P. H., & Tung, V. W. S. (2019). Examining the effects of robotic service on brand
functional aspects but also the social and emotional dimensions of experience: The moderating role of hotel segment. Journal of Travel & Tourism
human-robot interactions. These findings could have significant impli­ Marketing, 36(4), 458–468.
Chen, Q., Anders, S., & An, H. (2013). Measuring consumer resistance to a new food
cations for the coffee and food robot industries. A limitation of this study technology: A choice experiment in meat packaging. Food Quality and Preference, 28
arose from the recruitment method, which relied on a smartphone (2), 419–428.
application. Consequently, over 80% of the participants fell within the Cho, I. S. (2022, April). “Our chefs are robots”…From cooking to delivery, the evolution
of robots. Insight 42, Retrieved from https://www.tech42.co.kr/%EC%A0%80%ED
20–30 age range. As a result, the findings of this study may not be fully %9D%AC-%EC%85%B0%ED%94%84%EB%8A%94-%EB%A1%9C%EB%B4%87%
representative of all age groups. Further research is needed to explore EC%9E%85%EB%8B%88%EB%8B%A4-%EC%9A%94%EB%A6%AC%EC%97%90%
the impact of demographic factors on the acceptance of robot cooking as EC%84%9C-%EB%B0%B0%EB%8B%AC%EA%B9%8C%EC%A7%80-%EB%A1%
9C/. Accessed January 26, 2023.
well as for other food items that involve more complicated cooking Choi, Y. (2022, February). ’Cooking robot’ to replace cooking work in military units to be
processes, such as frying chicken and rice. introduced. Hypebeast, Retrieved from https://hypebeast.kr/2022/2/kitchen-pilice
Ethical statement: All participants voluntarily participated and -cooking-robot-military-camp-introduction. Accessed January 26, 2023.
Coppola, A., Verneau, F., & Caracciolo, F. (2014). Neophobia in food consumption: An
submitted informed consent forms before study engagement. Ethical empirical application of the FTNS scale in southern Italy. Italian Journal of Food
approval for the involvement of human subjects in this study was Science, 26(1).
granted by the Institutional Review Board (IRB) of the Korea Food Cotter, A. R., Batali, M. E., Ristenpart, W. D., & Guinard, J. X. (2021). Consumer
preferences for black coffee are spread over a wide range of brew strengths and
Research Institute (Approval number KFRI 2022–08-004–001).
extraction yields. Journal of Food Science, 86(1), 194–205. https://doi.org/10.1111/
1750-3841.15561
CRediT authorship contribution statement Cox, D., & Evans, G. (2008). Construction and validation of a psychometric scale to
measure consumers’ fears of novel food technologies: The food technology
neophobia scale. Food Quality and Preference, 19(8), 704–710.
Seyeong Park: Methodology, Formal analysis, Investigation, Evans, G., Kermarrec, C., Sable, T., & Cox, D. (2010). Reliability and predictive validity
Writing – original draft. Min Kyung Park: Methodology, Formal anal­ of the Food Technology Neophobia Scale. Appetite, 54(2), 390–393.
Feng, X., Khemacheevakul, K., De León Siller, S., Wolodko, J., & Wismer, W. (2022).
ysis, Investigation, Writing – original draft. JeongAe Heo: Investigation, Effect of labelling and information on consumer perception of foods presented as 3D
Resources. Ji-sun Hwang: Investigation, Resources. Sungjae Hwang: printed. Foods, 11(6), 809.
Conceptualization, Resources. Daekwang Kim: Conceptualization, Re­ Fusté-Forné, F. (2021). Robot chefs in gastronomy tourism: What’s on the menu? Tourism
Management Perspectives, 37, Article 100774.
sources. Seo-Jin Chung: Writing – review & editing. Han Sub Kwak:
Heidmeier, A. K., & Teuber, R. (2022). Acceptance of in vitro meat and the role of food
Conceptualization, Methodology, Supervision, Project administration, technology neophobia, dietary patterns and information–empirical evidence for
Funding acquisition, Writing – review & editing. Germany. British Food Journal, ahead-of-print.. https://doi.org/10.1108/BFJ-03-
2022-0244
Hwang, G. C. (2022, September). From coffee delivery to the way guidance, ’Advanced
robot’ navigating Incheon International Airport. Dong-A Daily News, from https://
Declaration of Competing Interest www.donga.com/news/article/all/20220926/115664647/1. Accessed January 26,
2023.
The authors declare that they have no known competing financial Ivanov, S. H., & Webster, C. (2017). Adoption of robots, artificial intelligence and service
automation by travel, tourism and hospitality companies–a cost-benefit analysis. Tourism
interests or personal relationships that could have appeared to influence
and Hospitality Companies–A Cost-Benefit Analysis.
the work reported in this paper. Kattara, H. S., & El-Said, O. A. (2013). Customers’ preferences for new technology-based
self-services versus human interaction services in hotels. Tourism and Hospitality
Data availability Research, 13(2), 67–82.
Kim, M. J., Suh, S. M., Kim, S. Y., Qin, P., Kim, H. R., & Kim, H. Y. (2020). Development
of a real-time PCR assay for the detection of donkey (Equus asinus) meat in meat
Data will be made available on request. mixtures treated under different processing conditions. Foods, 9(2), 130.
Kim, S. S., Kim, J., Badu-Baiden, F., Giroux, M., & Choi, Y. (2021a). Preference for robot
service or human service in hotels? Impacts of the COVID-19 pandemic. International
Acknowledgement Journal of Hospitality Management, 93, Article 102795.
Kim, Y. J., Park, J. S., & Jeon, H. M. (2021b). Experiential value, satisfaction, brand love,
and brand loyalty toward robot barista coffee shop: The moderating effect of
This research was supported by the Main Research Program (Grant
generation. Sustainability, 13(21), 12029.
number: E0211300) of the Korea Food Research Institute (KFRI) funded
by the Ministry of Science and ICT, Republic of Korea.

8
S. Park et al. Food Research International 172 (2023) 113119

Krings, V. C., Dhont, K., & Hodson, G. (2022). Food technology neophobia as a Park, S. J. (2019, October). People of delivery launches pilot service for indoor delivery
psychological barrier to clean meat acceptance. Food Quality and Preference, 96, robot. YTN, Retrieved from https://www.ytn.co.kr/_ln/0102_
Article 104409. 201910090010319087. Accessed January 26, 2023.
Kuang, L., Burgess, B., Cuite, C. L., Tepper, B. J., & Hallman, W. K. (2020). Sensory Schouteten, J. J., Gellynck, X., & Slabbinck, H. (2019). Influence of organic labels on
acceptability and willingness to buy foods presented as having benefits achieved consumer’s flavor perception and emotional profiling: Comparison between a central
through the use of nanotechnology. Food Quality and Preference, 83, Article 103922. location test and home-use-test. Food Research International, 116, 1000–1009.
Labbe, D., Ferrage, A., Rytz, A., Pace, J., & Martin, N. (2015). Pleasantness, emotions and Seo, H. S., Lee, S. Y., & Hwang, I. (2009). Development of sensory attribute pool of
perceptions induced by coffee beverage experience depend on the consumption brewed coffee. Journal of Sensory Studies, 24(1), 111–132.
motivation (hedonic or utilitarian). Food Quality and Preference, 44, 56–61. Seyitoğlu, F., & Ivanov, S. (2022). Understanding the robotic restaurant experience: A
Laureati, M., Conte, A., Padalino, L., Del Nobile, M. A., & Pagliarini, E. (2016). Effect of multiple case study. Journal of Tourism Futures, 8(1), 55–72.
fiber information on consumer’s expectation and liking of wheat bran enriched Seyitoğlu, F., Ivanov, S., Atsız, O., & Çifçi, İ. (2021). Robots as restaurant employees-A
pasta. Journal of Sensory Studies, 31(4), 348–359. double-barrelled detective story. Technology in Society, 67, Article 101779.
Lee, K. H., Hwang, K. H., Kim, M., & Cho, M. (2021). 3D printed food attributes and their Stock, R. M., & Merkle, M. (2018). Can humanoid service robots perform better than
roles within the value-attitude-behavior model: Moderating effects of food service employees? A comparison of innovative behavior cues. Proceedings of the 51st
neophobia and food technology neophobia. Journal of Hospitality and Tourism Hawaii International Conference on System Sciences.
Management, 48, 46–54. Sung, H. J., & Jeon, H. M. (2020). Untact: Customer’s acceptance intention toward robot
Lee, S., & Ham, S. (2021). Food service industry in the era of COVID-19: Trends and barista in coffee shop. Sustainability, 12(20), 8598.
research implications. Nutrition Research and Practice, 15(Suppl 1), S22–S31. Verneau, F., Caracciolo, F., Coppola, A., & Lombardi, P. (2014). Consumer fears and
Matin, A. H., Goddard, E., Vandermoere, F., Blanchemanche, S., Bieberstein, A., familiarity of processed food. The value of information provided by the FTNS.
Marette, S., & Roosen, J. (2012). Do environmental attitudes and food technology Appetite, 73, 140–146.
neophobia affect perceptions of the benefits of nanotechnology? International Journal Vidigal, M. C., Minim, V. P., Simiqueli, A. A., Souza, P. H., Balbino, D. F., & Minim, L. A.
of Consumer Studies, 36(2), 149–157. (2015). Food technology neophobia and consumer attitudes toward foods produced
Nestrud, M. A., Meiselman, H. L., King, S. C., Lesher, L. L., & Cardello, A. V. (2016). by new and conventional technologies: A case study in Brazil. LWT-Food Science and
Development of EsSense25, a shorter version of the EsSense Profile®. Food Quality Technology, 60(2), 832–840.
and Preference, 48, 107–117. Villegas, B., Carbonell, I., & Costell, E. (2008). Effects of product information and
Niimi, J., Collier, E. S., Oberrauter, L.-M., Sörensen, V., Norman, C., Normann, A., consumer attitudes on responses to milk and soybean vanilla beverages. Journal of
Bendtsen, M., & Bergman, P. (2022). Sample discrimination through profiling with the Science of Food and Agriculture, 88(14), 2426–2434.
rate all that apply (RATA) using consumers is similar between home use test (HUT) World Coffee Research (2017). Sensory lexicon – unabridged definition and references.
and central location test (CLT). Food Quality and Preference, 95, Article 104377. file:///C:/Users/User/Downloads/20170622_WCR_Sensory_Lexicon_2-0-1%20(1).
Nozawa, C., Togawa, T., Velasco, C., & Motoki, K. (2022). Consumer responses to the use pdf. Accessed July 13, 2022.
of artificial intelligence in luxury and non-luxury restaurants. Food Quality and Xiao, C., & Zhao, L. (2022). Robotic chef versus human chef: The effects of
Preference, 96, Article 104436. anthropomorphism, novel cues, and cooking difficulty level on food quality
Park, H. S. (2022, July). “Have a coffee made by a robot.” A robot coffee shop appears at prediction. International Journal of Social Robotics, 14(7), 1697–1710.
a highway rest area. NEWSIS, Retrieved from https://mobile.newsis.com/view. Zhu, D. H., & Chang, Y. P. (2020). Robot with humanoid hands cooks food better? Effect
html?ar_id=NISX20220711_0001938997#_PA. Accessed May 26, 2023. of robotic chef anthropomorphism on food quality prediction. International Journal of
Contemporary Hospitality Management, 32(3), 1367–1383.

You might also like