1. Introduction
Technology has grown and expanded in society, closely followed by educational uses. This growth has offered teachers more tools to assist in reaching their goals of helping students learn. However, the expansion of different types of technology has forced teachers and students to be prepared for constant technological adaptation (Veiga & Andrade, 2021). Some technologies flourish and become adopted by many, while others fail (Rogers, 1995). Therefore, teachers need to determine which technologies the students want or need to use to increase the skill being taught.
One innovation that has been used since ancient times concerning learning a foreign language is translators. Translators have been in use for centuries and have been used to learn about other cultures and languages, possibly dating back to Assyrian-Mesopotamian bilingual inscriptions (3000 BC) to the Rosetta Stone (196 B. C.) (Kelly, 1995). Communication was the focus of these forms of translation. Translators moved into the classroom with the Grammar-Translation Method of teaching, first published in 1845 (Richards & Rogers, 2014). At this time, students carried bilingual dictionaries to class to help with translation. Over time, the technology of translators has changed from tablets and rocks to books, and now the technology most used for translation is smartphone-based translators.
Smartphone translation is still in its early stages of development as a part of the 4th Industrial Revolution. The EFL classroom needs to reflect this by adopting new technologies that can be used for real-life skills (Alakrash & Razak, 2020; Ningsih, Suherdi, & Purnawarman, 2022). Other reasons for possible issues related to technology adoption in the classroom were teacher readiness and willingness to use technology, infrastructure issues and possible school policies (Ningsih et al., 2022). Teacher attitudes and self-perceived technology competence were major factors in adopting 4th Industrial Revolution skills in the EFL classroom (Wen & Kim Hua, 2020). Kit and Ganapathy (2019) added that teachers might have trouble staying on top of the latest technologies due to a lack of time and awareness. Other issues with instructors including smartphone translators into the EFL classroom had some other issues. First, instructors tend to use machine translator (MT) tools with less frequency than students (Ata & Debreli, 2021). Ata and Debreli (2021) also found that instructors felt that how the MTs were used was essential to their ethicality when used on English assignments. The findings reported that the passage length was related to the instructor’s belief in unethicality, with the longer passages being the most unethical (Ata & Debreli, 2021). Jolley and Maimone (2022) found that many instructors had issues using MT in the classroom, especially the possible unethical usage. However, they concluded that some instructors are starting to see their use as inevitable, bordering on useful. Liu, Kwok, Liu, and Cheung (2022) even believe that MT should be added to the curriculum to help students recognize their best use.
However, teachers are only one side of the classroom. Students and their attitudes and perceptions of the use of technology in the classroom, and more specifically smartphone translation, should be considered. Jones, Richards, Cho, and Lee (2019) found that while students liked to use technology for everyday life, they were not favorable to classroom technologies. Kim and Han (2023) found that students felt that they had improved their post-editing skills using Google Translate as part of the EFL writing classroom. Dennis (2019) also found a generally positive attitude towards machine translation with a few reservations. This study attempts to understand how student perceptions of smartphone translators relate to their usage in the EFL writing classroom.
2. Literature Review
2.1. Use of Smartphone Translators in the EFL Classroom
Smartphone translators have made machine translation (MT), which used to be bound to computers, more accessible to many students. In the Korean context, Google Translate, Papago, and Kakao i have become the most common machine translators used on smartphones in the EFL classroom (Koh, 2022). Kim and Han (2023) determined that the teacher is integral to student success in incorporating MT into the EFL classroom. Using smartphone translators in the EFL writing classroom has several advantages and disadvantages.
The first advantage of smartphone translators is that the output quality is good, especially with critical thinking about the results. Groves and Mundt (2015) found that the translation level by Google Translate was approaching a 6.0 IELTS and improving, but the output quality depended on the input language. Lee and Lee (2021) concluded that with self-judgment, students could use MT to produce high-quality writing. Dennis (2019) determined that students used MT to look up words and phrases as they needed help. Chung (2020) found that if a student had a threshold level of English proficiency, they could use MT to create good writing by being critical of the translator’s output. However, Garcia and Pena (2011) found that MT helped lower-proficiency students.
Second, smartphone translators are easy for language learners to use. Chandra and Yuyan (2018) explored how students used MT and found that a majority used it for looking up words (68.1%) or phrases (20.2%) instead of sentences, spelling, or grammar. They preferred to use Google Translate to a paper bilingual dictionary because it was easier and faster. Almusharraf and Bailey (2023) found that MT was easy to use, indicating how user-friendly they are. Jolley and Maimone (2022) concluded that students were aware of the limitations of MT but continued to use them due to the ease of use and the perceived usefulness of the device, which could eventually lead to an increase in a student’s metalinguistic knowledge.
Finally, smartphone translators allowed learners to understand better the language and culture of the L1 and L2. Chung (2020) determined that using MT paired with classroom activities helped the students better understand the differences and similarities between the two languages. Almusharraf and Bailey (2023) found that MT allowed students to explore the differences between culture and language using MT. Kim and Han (2023) concluded that with other educational resources, MT allowed students to discover the nuance between the languages.
However, there were also some negative aspects of smartphone translators. The first issue relates to whether the students are shortcutting the learning process using MT. Benda (2014) was concerned that students using MT were more interested in finishing the task than in the communicative process. Dennis (2019) concurred that one of the liabilities of using MT was an overreliance on them. Dennis (2019) continued that students were worried about long-term vocabulary retention when this occurred. Byun (2022) also concluded that students used MT when they did not want to think through the answer.
The second possible issue of using smartphone translators was possible concerns with the output quality. Chandra and Yuyan (2018) found that some interviewees felt that some translations needed to be corrected, especially with more extended groupings of words. This finding is supported by Koh (2022), who compared human translation and MT for a movie script. MT had an error rate between 22% and 40% for the three MT tested: Google Translate, Papago, and Kakao i. Lee and Lee (2021) found that the highest-level students in the group chose not to use some of the translations. The students attempted to correct them by rephrasing or searching for alternative English expressions.
The third issue of using a smartphone translator relates to the blind reliance of lower-level students on the results. This decision process noted by Lee and Lee (2021) indicated that the higher the student’s level, the more they could discern the difference between good and bad output quality. In one case, Lee and Lee (2021, p. 45) reported, ”In this case, I preferred to use the output from Papago because I did not know better expressions.” Chung (2020) found that the results of MT were directly proportional to a student’s proficiency level because the structure of the MT output did not limit the more advanced students. Chung (2020) found that the lower-level students tended to accept the output from the MT because of a lack of lexical knowledge. Lee and Lee (2021) concurred by offering the possible solution of EFL teachers offering instruction related to the strengths and weaknesses of MT output and linguistic differences between the two languages.
2.2. Technology Acceptance Model
Technology has become ubiquitous in society, and the classroom is a part of this spread. Technology applications in the classroom continue to grow and change with society. Gauging student acceptance of varied technologies is one tool to increase the utilization of technology. One method to determine the learners’ feelings is the Technology Acceptance Model (TAM). The TAM was first developed by Davis (1985) and then expanded into TAM2 by Venkatesh and Davis (2000). The TAM uses a series of factors within a model to predict student acceptance of a specific type of technology.
The original TAM model was based upon a user’s perceived usefulness (PU) and perceived ease of use (EoU) positively affecting the end usage of a given technology (Davis, 1989; Davis et al., 1989; Venkatesh & Davis, 2000). First, the perceived usefulness of a technology indicates how much aid a technology gives a worker or learner in completing a task. The attitude will decrease if the learner does not believe the technology will help complete the task. Next, the perceived ease of use is the amount of work the user will exert to complete the task. Even if the user understands that the task can be completed well with technology, it must take little effort. The learner needs to believe that it is easy to use because if a user does not think that, they are less apt to have a positive attitude towards it (Venkatesh & Davis, 2000). These factors are in most TAM models for using technology amongst a population.
The original TAM was altered to add more external social factors to the equation, creating TAM2 (Venkatesh & Davis, 2000). These external factors were created to add more items that could influence PU than just EoU. TAM2 added the subjective norm (SN), voluntariness (V), output quality (OQ), and results demonstrability (RD) as factors in a user’s PU of a given technology. Some TAM models include the category of external factors (EF). EF combines all or some of the factors added to Venkatesh and Davis’s (2000) TAM2 (Liu & Ma, 2023). While this might work for some technology uses, multiple facets of the external factors pertain to student MT use. MT has mixed reports related to OQ (Chandra & Yuyan, 2018; Dennis, 2019; Groves & Mundt, 2015; Koh, 2022; Lee & Lee, 2021) and RD (Almusharraf & Bailey, 2023; Chung, 2020; Kim & Han, 2023; Lee & Lee, 2021). Therefore, they should be separated within the TAM structure in this case.
The first social factor was SN. The subjective norm was defined by Fishbein and Ajzen (1975, p. 302) as a ”person’s perception that most people who are important to him think he should or should not perform the behavior in question.” The reason for the inclusion of SN in TAM2 is that learners might not think technology is helpful by themselves, but if others around them do, they could be more prone to add to their perceived usefulness (Venkatesh & Davis, 2000). SN’s inclusion in the survey could determine if peer usage affects the students’ attitudes toward MT.
Next, the concept of voluntariness (V) was also added to the TAM2. The concept that there is a perception that the decision to use technology belongs to the user is the definition of V (Agarwal & Prasad, 1997; Venkatesh & Davis, 2000). Hartwick and Barki (1994) supposed that mandatory usage led to more extensive changes in the social factors. However, if all the learners in the class use the technology, it could feel required to be able to compete with the others. As some studies relate to incorporating MT into the classroom (Chung, 2020; Dennis, 2019; Kim & Han, 2023; Lee & Lee, 2021) this factor should be included to determine its role in the model.
Output quality (OQ) is the degree to which ”people will take into consideration how well they perform those tasks” (Venkatesh & Davis, 2000, p. 191) using the technology. Therefore, the perception that the output is higher quality adds to the PU, making it a distinct factor in the model. With output quality being one of the main concerns for the use of smartphone translators (Chandra & Yuyan, 2018; Koh, 2022; and Lee & Lee, 2020), the concept of OQ should be included in the TAM model’s structure to determine if the learners in the study had the same type of concerns.
Results demonstrability (RD) is the final external factor in the TAM2. RD is the extent to which the technology aids the learner to reach a goal and understand how the technology works. RD is also found in the literature, with some studies voicing a favorable opinion (Almusharraf & Bailey, 2023; Chung, 2020; Kim & Han, 2023), while others sounded some warnings about RD (Chung, 2020; Lee & Lee, 2021).
For this survey, the TAM was broken into three distinctive stages. The first stage determines which external factors, including EoU, SN, V, OQ, and RD, affect a learner’s PU of using a smartphone translator in an EFL writing class. The second part of the model explores how EoU, SN, and PU affect the learner’s intention to use (IU). The final phase determines IU’s relationship with the user’s usage behavior (UB) (see Figure 1).
Based on the structure of the TAM model for EFL smartphone translator usage, the following three research questions will be explored:
RQ1. How do SN, V, OQ, RD, and EoU relate to the learner’s perceived usefulness (PU) related to the use of smartphone translators in the EFL classroom?
RQ2. How do SN, PU, and EoU relate to EFL students’ intention to use (IU) smartphone translators in the classroom?
RQ3. How does IU relate to EFL students’ smartphone translator usage behavior (UB)?
3. Methods
3.1. Research Subjects
The participants consisted of 194 students enrolled in a required EFL course at a medium-sized university in South Korea. Some of the students failed to answer all the questions in the survey and were excluded listwise from the results. This deletion led to only 181 respondents being used for data analysis. The survey asked questions to determine student perceptions of using AI-based smartphone translators based on the Technology Acceptance Model (TAM). The respondents completed the survey during the 2022 and 2023 Spring terms. Most students reported that they were in their first year of university, with 156 choosing that option. Nine second-year and four third-year students took the survey. There were 22 fourth-year students and three did not report their year at university. The participants represented a wide range of majors, with business being the largest at 69. The study also had 54 engineering majors and 13 humanities majors. Ten students replied that they were social science majors. Art, education, and law students represented eight, seven, and four responses. Twenty-nine students chose not to report their major. The students were enrolled in intermediate-level English as a foreign language writing course. A TOEIC-style test was used before student enrollment for placement into the level - students with scores of 401 to 700.
3.2. Research Tools
The questions were put into a Google Form, and the link was shared with the students through the university’s online system at the end of the semester. There was a notice that all participation was voluntary and that the material collected would not include any identifying information. The students were asked to respond to the 22 questions based on the Venkatesh and Davis (2000) Technology Acceptance Model (see Table 1). Each question was a four-point Likert-type scale, with four equating to strongly agree and one representing strongly disagree. The questions were posed in English and Korean to ensure all the students could understand them.
Table 1
Items from Questionnaire, Cronbach’s Alpha, and Factor Loadings
3.3. Research Procedures
The class was a three-credit hour class that met twice a week over a 15-week semester. The focus of the class was EFL academic writing. One of the thirty classes was dedicated to talking about how to get the best results possible when using a smartphone translator and possible pitfalls. Students completed writing projects and in-class assignments. During the in-class assignments, the teacher circulated around the class and helped the students with their writing. This help included an assessment of how well the smartphone translators’ output met the goals of the course. In 2022, there were limited COVID precautions in place including the wearing of masks and some social distancing. However, the classes met in the room, and the teacher could interact with the students. In 2023, the course management was not affected to the same degree by COVID.
3.4. Data Analysis Methods
The results of the survey were exported from Google Forms to IBM SPSS Statistics (Version 27) for statistical analysis. First, the descriptive statistics were determined. If participants did not answer all the questions, they were deleted listwise from the data. The next stage of statistics was to check the internal reliability and feasibility of performing a factor reduction using the data. After reducing the factors, the factors were analyzed using a linear reduction to identify how the elements interacted within the TAM model.
3.4.1. Statistical Assumptions for Factor Reduction
To assess if the collected data is appropriate for factor analysis, the Kaiser-Meyer-Olkin measure of sampling adequacy (KMO) and Bartlett’s test of sphericity (BTS) are necessary (Beavers et al., 2013). The data returned a KMO number of .851, which, according to Dziuban and Shirkey (1974), indicates that the sampling could be considered meritorious. The BTS was found to be significant χ2 (300) = 2486.845, p = < .004. According to Shrestha (2021), this result indicates that factor analysis can be performed.
To determine the items and number of factors to be included in the analysis, the load factor of each item and the percent of the total variance explained need to be determined. Peterson (2000) performed a meta-analysis on 401 studies that utilized factor analysis and posited that 56.6% of the variance explained and factor loads of .320 are necessary for items to be included. Using a baseline of 1.000 eigenvalues, the total variance explained was 64.97%. There were eight factors which had more than 1.000 eigenvalues. Using principle component analysis as a factor extraction method, the Varimax rotation method with Kaiser normalization was employed to break the items into the eight factors (see Table 2). The 22 items in the study had factor loads ranging from .503 to .823 (see Table 1). Therefore, all the items were appropriate for the factor analysis.
Table 2
Factor Breakdown of Items Using Principle Component Analysis with Varimax Rotation and Kaiser Normalization
The Cronbach’s alpha was determined for the entire study and each factor to determine the internal consistency each contained. Tavakol and Dennick (2011) wrote that the Cronbach’s alpha should be between the range of .700 and .950 but warned that it could be lower if there are a low number of questions. The overall survey had a Cronbach’s alpha of .886. The first factor, UB, had a Cronbach’s alpha of .804, while IU was at .907. PU and EoU had Cronbach’s alphas of .889 and .781, respectively. The fifth factor, SN, was at .883. The sixth and seventh factors, V and OQ, had slightly low Cronbach’s Alphas of .678 and .626. These slightly lower numbers could be due to the number of questions in the factor. Finally, results demonstrability had a Cronbach’s alpha of .784 (see Table 1). Therefore, all the assumptions to perform a factor analysis have been met.
3.4.2. Data Analysis for Factors Affecting Perceived Usefulness
For the first research question, the dependent variable was perceived usefulness (PU), and the predictive variables were ease of use (EoU), subjective norm (SN), results demonstrability (RD), voluntariness (V), and output quality (OQ). a multiple regression was performed to determine how the predictors affect the perceived usefulness. An ANOVA indicated the significance of the predictors to perceived usefulness in the TAM model. The adjusted R2 determined what percentage of the results the model predicts. The standardized coefficient beta determines the amount each of the factors affects PU.
3.4.3. Data Analysis for Factors Affecting Intention to Use
The second research question has the dependent variable intention to use (IU), which was predicted by EoU, SN, and PU as described by Davis and Venkatesh (2000). Once the factors are developed, a multiple regression ascertained the relationship between the variables. An ANOVA determined that the model had a significant effect. The multiple regression indicated the amount of effect the model contained through the adjusted R2. Finally, the coefficients determined in the multiple regression indicated the amount each predictor affected IU.
3.4.4. Data Analysis for Factor Affecting Usage Behavior
Finally, the third research question determines how the predictive variable IU leads to the dependent variable usage behavior (UB). A multiple regression was run to see if there is a significant effect within the model. The adjusted R2 indicated the size of the effect with an ANOVA performed to check whether the results are significant. The standardized coefficient beta of the model indicated the extent that IU affects UB.
4. Results
This section reports the results of the survey. First, the descriptive statistics for each question of the survey are explained. Next, the assumptions that indicate if a factor analysis can be performed are described. Finally, the standard multiple regressions to determine how the elements of TAM interact are reported. The multiple regressions move from the external factors through perceived usefulness to the intention to use and finally towards the reported usage behavior of EFL student use of smartphone translators.
4.1. Factors Affecting Perceived Usefulness
RQ1. How do SN, V, OQ, RD, and EoU relate to the learner’s perceived usefulness (PU) related to the use of smartphone translators in the EFL classroom?
First, perceived usefulness (PU) had four items straddling the agree-line with mean ranges between 2.97 to 3.07. The second part of the survey pertained to perceived ease of use (EoU) with four questions with means between 2.76 and 3.26. The third factor measured in the survey was subjective norm (SN). There were two items with a mean of 2.38, indicating a trend towards disagreeing with the survey items. Fourth, voluntariness (V) was addressed with three points with a range of means between 3.27 and 3.41. Finally, output quality’s (OQ) two questions had means between agree and disagree with means of 2.11 and 2.61. The final section results demonstrability (RD), had means of 2.89, 3.04, and 3.15 (see Table 3).
Table 3
Descriptive Statistics for PU, EoU, SN, V, OQ, and RD
Most of the means for the items, 11 out of 18, were between the agree and strongly agree responses. The highest mean was 3.41 on Item 16. My teacher does not require me to use a smartphone translator. Seven of the items had means under three. Of these items, four shaded towards agree with scores over 2.50, and three trended towards disagree, between 2.00 and 2.50. The lowest of all responses, Item 19. I have no problem with the quality of a smartphone translator’s output, had a mean of 2.11, indicating some problems for the students with getting the best results from smartphone translators. The standard deviations for all the items in the survey were between .631 and .855 (see Table 3).
In the first stage, perceived ease of use (EoU), subjective norm (SN), voluntariness (V), output quality (OQ), and result demonstrability (RD) were the factors of PU in the model. Within this model, all of the items were correlated in a statistically significant manner. The numbers all fall between .30 and .50, indicating a medium effect size (Hemphill, 2003). All but two of the inter-item correlations are also statistically significant (see Table 4). These relationships indicate a proper sample size for multiple regression (Amjad et al., 2020).
Table 4
Standard Multiple Regression to Predict PU from EoU, SN, V, OQ, and RD
| R | Adjusted R2 | F | df | Sig. | |
|---|---|---|---|---|---|
| Model Summary | .605 | .34 | 16.47 | 6,175 | .004** |
The overall regression utilizing the five predictors was statistically significant at an adjusted R2 = .34, F (6, 175) = 16.47, p = .004 (see Table 4). The result describes a medium effect size of 34% of the regression variance. Three factors described a statistically significant relationship with a positive slope. The positive slope indicates that as the predictor increases, PU also increases. SN had the largest standardized coefficient in the model with β = .39, t (181) = 5.73, p = .004. V was the next largest standardized beta with β = .19, t (181) = 2.50, p = .013. OQ had a statistically significant effect at β = .15, t (181) = 2.03, p = .044. EoU and RD did not have a statistically significant effect concerning PU (see Table 5).
Table 5
Correlations and Beta Coefficients for PU from EoU, SN, V, OQ, and RD
| PU | EoU | SN | V | OQ | β | t | Sig. | |
|---|---|---|---|---|---|---|---|---|
| Perceived Ease of Use (EoU) | .335** | .08 | 1.03 | .305 | ||||
|
|
||||||||
| Subjective Norm (SN) | .494** | .185* | .39 | 5.73 | .004** | |||
|
|
||||||||
| Voluntariness (V) | .319** | .460** | .054 | .19 | 2.50 | .013* | ||
|
|
||||||||
| Output Quality (OQ) | .393** | .385** | .418** | .124 | .15 | 2.03 | .044* | |
|
|
||||||||
| Results Demonstrability (RD) | .343** | .467** | .170* | .522** | .236** | .10 | 1.34 | .181 |
4.2. Factors Affecting Intention to Use
RQ2. How do SN, PU, and EoU relate to EFL students’ intention to use (IU) smartphone translators in the classroom?
IU was the only new factor to answer research question 2, with SN, PU, and EoU already being described. IU had the highest average, with the two items having the highest and third highest means at 3.41 and 3.40, respectively. These averages indicate that the students were between agreeing and strongly agreeing to use smartphones in the EFL writing classroom. Eight out of 12 items had a mean between agree and strongly agree. Of the four between agree and disagree responses, two shaded positively, and two shaded negatively. Both items shaded towards disagree were in SN at 2.38 (see Table 6).
Table 6
Descriptive Statistics for IU, PU, EoU, and SN
Next, a standardized regression was performed to determine if SN, PU, and EoU were statistically related to IU. This stage of the model had an adjusted R2 = .32, F (3, 178) = 28.77, p = .004 (see Table 7). All the correlations between IU and the predictors were statistically significant at p < .01. PU and EoU displayed correlations between .30 and .50, indicating a medium-sized effect on the model. SN also showed a statistically significant correlation, but only in the low effect size category of .10 to .30 (Hemphill, 2003). All the inter-item correlations were also statistically significant, intimating that the test was justified. Two predictors were statistically significant in the model, with PU displaying a β = .41, t (181) = 5.55, p = .004, and EoU with a β = .32, t (181) = 4.88, p = .004. SN was not significantly associated with IU (see Table 8).
Table 7
Standard Multiple Regression to Predict PU from EoU, SN, V, OQ, and RD
| R | Adjusted R2 | F | df | Sig. | |
|---|---|---|---|---|---|
| Model Summary | .57 | .32 | 28.77 | 3,178 | .004** |
4.3. Factor Affecting Usage Behavior
RQ3. How does IU relate to EFL students’ smartphone translator usage behavior (UB)?
The first factor was usage behavior (UB), which had two questions in the survey. The means for the items were 2.28 and 2.52. Both items were between the agree and disagree response levels. The second factor, intention to use (IU), consisted of two items with nearly identical means of 3.41 and 3.40. Students agreed with IU by responding with the highest average mean of any faction. The responses for UB were both between agree and disagree, with one being close to neutral and one tending more towards disagreement (see Table 9).
Table 9
Descriptive Statistics for IU and UB
Finally, the researcher ran a standard regression to determine how IU influenced UB. This section of the model had a statistically significant adjusted R2 = .044, F (1, 180) = 9.77, p = .002 (see Table 10). This result indicated that only 4% of the variance of the model can be explained by this relationship. IU had a statistically significant standardized coefficient β = .22, t (181) = 3.13, p =.002 (see Table 11).
Table 10
Standard Multiple Regression to Predict UB from IU
| R | Adjusted R2 | F | df | Sig. | |
|---|---|---|---|---|---|
| Model Summary | .222 | .044 | 9.77 | 1,180 | .002** |
Table 11
Standardized Coefficients Beta to Predict UB from IU
| β | t | Sig. | |
|---|---|---|---|
| Perceived Usefulness (PU) | .22 | 3.13 | .002** |
5. Discussion
The first research question attempts to determine what kind of relationship, if any, there is with the subjective norm (SN), voluntariness (V), output quality (OQ), results demonstrability (RD), and perceived ease of use (EoU) affecting the perceived usefulness (PU) of student use of smartphone translator in the EFL classroom. The results show statistically significant relationships between three of these factors and a student’s PU. The effect variables for SN, V, and OQ were all positive, indicating that as the student’s attitude towards smartphone translators increased, their perceived usefulness also increased. RD and EoU did not have a statistically significant effect on PU. This section of the model had an adjusted R2 = .34, indicating that 34% of the model is explained with these variables (see Figure 2).
The first factor in this section of the model is SN. SN had the largest statistically significant, positive effect on PU. With a β = .39, there was almost a medium effect size, double what the other significant factors displayed. The positive effect size indicates that students who rated SN high tended also to rate PU high. With a mean of 2.38 for each of the questions related to SN, the participants tended towards disagreeing with the points that made up SN. This reaction could indicate that most students did not feel pressure from people to use smartphone translators. However, the students with high results in SN perceived smartphone translators as more useful.
The next factor in the model was voluntariness (V). The means of the three items related to V were between 3.27 and 3.41, all between agreeing and strongly agreeing points of the scale. V had a statistically significant effect on PU. The β = .19 indicates a small effect size on the model. The high means and statistically significant β confirm that students believed that the use of smartphone translators was their choice, which made them more valuable in the eyes of the participants.
The final factor that had a statistically significant effect on PU was output quality (OQ). With a β = .15, OQ has a small positive effect on PU. The means of the questions related to OQ were 2.61 and 2.11. The first answer shades towards agreeing but is very close to neutral. The second mean was the lowest of any question in the survey, being the closest to the disagree choice than any other question. This indicates that students did not fully trust the output generated as was suggested in the literature (Chandra & Yuyan, 2018; Koh, 2022; Lee & Lee, 2021). Even though the responses were low, the positive effect indicates that the PU also increases as the participant attitudes towards OQ increase.
Finally, two factors did not have a statistically significant effect on PU: results demonstrability (RD) and perceived ease of use (EoU). First, RD’s responses to the three questions straddled agree with a range of 2.89 to 3.15. Therefore, even though students felt they understood and could communicate the results, this did not directly affect how they perceived smartphone translator usefulness. EoU displayed similar means as RD with a range from 2.76 to 3.26, but it also did not display statistically significant results on the perceived usefulness.
Students were aware of pressure to use smartphone translators from those around them as indicated by the subjective norm (SN). This peer pressure led to students seeing the MT as more useful than they would have in a vacuum. Interestingly, despite peer pressure, the data showed that students felt that voluntariness (V) was positively statistically related to perceived usefulness (PU). This interesting dichotomy indicates that PU is built through peer pressure, not teacher pressure, to use smartphone translators. Students who had a sense of getting higher output quality (OQ) tended to also have a higher PU. However, OQ had low student means, so teachers who want to improve student perceptions of smartphone translators need to help students find ways to improve their competency. This coincides with the findings of Liu et al. (2022). Results demonstrability (RD) and perceived ease of use (EoU) did not statistically affect the model. Therefore, teachers should attempt to help students understand how a smartphone translator works.
The second research question related to the TAM investigates the relationship of SN, PU, and EoU with an EFL learner’s intentions to use (IU) smartphone translators. Two of the three factors in this part of the model have a statistically significant effect on the EFL learners IU: PU and EoU. Both PU and EoU had a positive β, indicating that as the student attitude increases, so does IU. However, SN did not statistically influence IU. This stage of the TAM resulted in an adjusted R2 = .32.
First, PU had a statistically significant effect on IU. The effect size was medium at β = .41. Therefore, when a student’s PU increases, IU does as well. The four questions that comprised PU had means between the values of 2.97 and 3.07. This range is very close to the survey’s response agree. Accordingly, the perceived usefulness is essential to an EFL learner’s intention to use. While SN did not have a direct statistically significant effect on IU, SN has a close to medium effect size on PU, so there is an indirect effect.
Finally, the EFL learner attitudes related to ease of use (EoU) have a positive statistically significant relationship to IU. The β = .32 indicates low to medium effect size. Therefore, students with a positive outlook towards EoU tend to have a stronger IU. The EFL learners in the study tended to agree with all four questions related to the ease of use with means of 3.26, 3.19, 3.09, and 2.76. Interestingly, EoU did not have a statistically significant effect on PU in the first part of the model but did with IU. This could be a sign that there was a sentiment that the ease of use did not necessarily make it useful, but the EFL learners would use it anyway. These results indicate that students are using smartphone translators because they are easy, not to learn as was found in the literature (Almusharraf & Bailey, 2023; Benda, 2014; Byun, 2022; Chandra & Yuyan, 2018; Dennis, 2019).
Perceived ease of use (EoU) did not statistically affect PU but was positively statistically related to the intention to use (IU). This indicates that students would use a translator even if they did not have a high PU. If teachers decide to utilize smartphone translators as classroom tools, they should work more to build on a student’s OQ and RD to establish better that EoU is a part of PU as well as IU. PU was also positively related to IU. According to the TAM model, SN could have influenced IU and PU, but it did not. This indicates that the students only perceive the usefulness of SN but that it does not continue to IU.
The third research question asked how IU related to EFL student usage behavior (UB) with smartphone translators. The adjusted R2 = .04 was statistically significant, but low indicating other forces tied to learner UB of smartphone translators not covered in this model. The β = .22 indicates a small effect size from IU to UB. The higher a student’s IU, the higher that UB will tend to be. IU had means that were tied for the highest in the survey, 3.41, and third at 3.40. This means placing the responses solidly between agree and strongly agree. However, UB had means between agree and disagree, 2.52, and trending towards disagree at 2.28. This disparity could be one of the reasons for the lower R2. It is interesting that the question Given that I have access to a smartphone translator, I intend to use it was so high (3.41), and At my school, usage of a smartphone translator is important had such a low score: 2.28. This could indicate that students did not feel they had to use a smartphone translator, but if they had one, the trend was to utilize it.
In the final part of the TAM model, IU had a positive statistically significant effect on usage behavior (UB). UB also had low means from the participants, indicating that students need to be sure about the importance and relevance of smartphone translators in the EFL classroom. Teachers could increase this by incorporating activities to build student proficiency in using smartphone translators. Based on the results of this study using the TAM, smartphone translators could be helpful tools to help students improve their EFL skills if activities can be formulated to increase the student’s attitude towards OQ and UB.
6. Conclusions and Suggestions
The present study explored student perceptions and attitudes toward smartphone translators using Venkatesh and Davis’s (2000) TAM model. First, the study analyzed which external factors affected the student’s perceived usefulness of smartphone translators. The second stage was an investigation of the factors that made a student intend to use smartphone translators. Finally, the model examined how the intention to use smartphone translators affected a student’s usage behavior.
The students perceived the problems with the output quality from using smartphone translators. Groves and Mundt (2015) rated Google Translate as the English level with the lowest score necessary to study abroad at 6.0 on the IELTS. Koh (2022) found a high error rate in all three smartphone translators commonly used in South Korea. Lee and Lee (2021) offered the solution of teaching students to use tools like looking up the expressions, searching for alternative meanings, and confirming the meaning. These tools allow the student to critically judge whether to use, edit, or discard the smartphone translator’s output. Chung (2020) found that a certain proficiency level in English was necessary for students to accomplish what Lee and Lee (2021) outlined. Based on this, if smartphone translators are used in the classroom, instructors should consider spending some of the time teaching critical thinking skills to determine if the output is usable.
The students perceived smartphone translators to be easy to use. However, since ease of use was statistically tied to intention to use but not perceived usefulness, students might use smartphone translators as a shortcut to completing the task instead of as a learning tool as was also found in the literature (Benda, 2014; Byun, 2022; Dennis, 2019). This result could lead to instructors avoiding including smartphone translators in the classroom. However, while completing homework, it is difficult for the instructor to stop a student from taking the easy way of mindlessly using a smartphone translator. Therefore, there is a gap between the students and the instructors. Instructors should consider how much they plan to use smartphone translators in the classroom and what activities can overcome this ”easy way out” behavior. One possible activity that Lee and Lee (2021) used was to have students use smartphone translators to post-edit a passage to find and correct errors.
Results demonstrability was not statistically tied to the model but had responses indicating that students thought they could explain the results a smartphone translator produced. This result indicates that students needed to connect the idea of being able to explain why a smartphone translator worked and their usefulness. The students’ level could have caused this gap in the proposed model. The highest-level students were found to be able to determine the difference between usable and flawed output (Chung, 2020; Lee & Lee, 2021). The students in the study were in an intermediate-level class and could have been too low to understand the differences and similarities between English and Korean. To help connect the two factors, linguistically and culturally, working with the students on the nuance between L1 and L2 is necessary (Almusharraf & Bailey, 2023; Chung, 2020; Kim & Han, 2023). Instructors should consider building more scaffolding into MT lessons to aid students in making proper critical decisions about the results.
Smartphone translators should be systematically added to EFL writing courses. This study showed that students intend to use MT, despite not trusting the output quality. Some form of critical and systematic use of smartphone translators would help the results the students produce in a way that also improves English proficiency.
There are some limitations of the study which bear further research. First, the students were all in the same level class. It would be interesting to see if the model changes for students at different levels. Second, there are factors beyond what is in the TAM that are affecting the model. Determining these factors would help to develop what teachers could do to help students get more out of smartphone translators through targeted activities. Next, this study was only given to students taking a writing course. It would be interesting to see how students studying other language skills, like reading, speaking, and listening, perceive the use of smartphone translators. Finally, the study was conducted without systematic smartphone translator use other than one class at the beginning of the semester and one-on-one in-class aid in their use.
PDF Links
PubReader
ePub Link
Full text via DOI
Download Citation
