Technology Acceptance in Smart Agriculture

Objective

Increasing populations and climate change have had a significant impact on global agriculture, emphasising the need for both increased food security and greater sustainability. The last decade has also witnessed an explosion of information and communication technologies (ICTs) that may significantly disrupt traditional agriculture, both positively and negatively. Smart agriculture has the potential to help in addressing global agricultural challenges, but success will be dependent on the degree to which new technologies are accepted by the agriculture community. It is therefore important to measure the acceptance of such technology. However, such measurement is difficult as no validated and established measure of technology acceptance in smart agriculture currently exists.

Acceptability can be defined from the perspective of the user i.e., the process that the user goes through in decision making to make sure that an object is suitable and accepted. Several formal scales and models currently exist to assess acceptance. For example, the Technology Acceptance Model (TAM) has been validated in different domains such as IT usefulness and ease of use, and IT adoption by employees. There were studies on a meta-analysis of 88 TAM studies in different areas of Information Technology. However, no such scale exists within the domain of smart agriculture. The purpose of this paper is to develop such a scale.

Work

Papers were identified by performing an online search using keywords (see Table 2) of journal articles and conference proceedings from the following databases: ACM Digital Library, Scopus, CABI and Business Source Complete (EBSCOhost). The PICO (Population Intervention Comparison and Outcome) acronym was adapted to describe the inclusion and exclusion criteria used for the identification of papers. Our focus was on the user studies in crop farming based on technology interventions (see Table 1).

Table 3 presents the paper selection process adapted to the PRISMA statement and the screening results independently for each database. Of the total of 16 articles that met the inclusion criteria, the research (n=2 each) was conducted in Australia, France, India and Indonesia. Others (n=1 each) were conducted in Brazil, China, Czech Republic, Germany, Italy, Philippines, Spain and Thailand. The articles were published between 2009 and 2020. The participant numbers ranged from 10 to 727, composed mainly of farmers and agronomists, but also including smaller groups such as lecturers, employees, researchers and students. Quantitative and qualitative research methodologies were used across these research studies.

The Theoretical Acceptance Model (TFA) illustrates acceptability from the user's point of view using 7 constructs. They are:

1. Affective Attitude which describes how users feel about the system

2. Burden which describes the perceived amount of effort needed by the user to use the system

3. Ethicality which describes the degree to which the system blends in with the user’s value system.

4. Intervention/System Coherence which describes the degree to which the user understands the system and its working

5. Opportunity Costs which captures what (and the degree to which), the user must relinquish in order to use the system

6. Perceived Effectiveness describes the degree to the users perceive the system as delivering anticipated results

7. Self Efficacy which describes if the users are sufficiently confident to make the necessary behaviour changes required by the system.

We mapped the shortlisted papers to the TFA as it would help us to capture all the elements (7 constructs) of technology acceptance. This can be clearly observed in Table 4 where the shortlisted articles mapped only with some of the constructs of TFA as it was only considered and/ evaluated as part of technology acceptance. Such a mapping can help us identify how technology acceptance is presently defined and captured in crop farming.

The results of the shortlisting of the literature are shown in the Appendix, which lists 198 scale items based on studies reported in these 16 papers. Unfortunately, most studies do not report on the scale construction, reliability, or validation. The exceptions are (Li et al. 2020; Caffaro et al. 2020; Mir and Padma 2020; Michels, Bonke, and Musshoff 2020), however, they all measured technology acceptance from different perspectives and have limitations.

Li et al. (2020) adapted the scale that was previously validated in the domain of mobile banking (Zhou, Lu, and Wang 2010) which produced cross-loadings and the items had to be removed but the authors’ have not revalidated them in the farming domain. Caffaro et al. (2020) adapted items from scales validated in agriculture and information technology with a combination of adoption factors from agriculture (Davis 1989; Adrian, Norwood, and Mask 2005; Kernecker et al. 2019; Pierpaoli et al. 2013; Unay Gailhard, Bavorová, and Pirscher 2015), which gave rise to model fit issues and the authors’ have not revalidated them in the farming domain. Michels, Bonke et al. (2020) adapted the scale items from information technology. However, a scale with good internal consistency has at least three items per factor to have at least three or four items with high loadings per factor which is not followed by the adapted scale where one factor has only two items and another factor has only one item where internal consistency cannot be checked. Mir and Padma (2020) have not clearly explained where the scale items were derived from. The authors’ provided the complete list of items (237) when contacted directly. These items were reduced to 105 after the reliability testing which is still not provided in the paper. However, the authors have not followed the advice of having at least three items per factor to have at least three or four items with high loadings per factor as there were nine factors with only two items each. Their results suggest that there may be cross-loadings between their scale factors as they found a high correlation between a few factors which were not eliminated. This might explain why eight factors had at least five or more items. The authors’ have not revalidated the 105 items scale in the farming domain.

Results

We reduced the 198 items listed in the Appendix in four steps. First, the duplicates were removed, merged with highly similar items and grouped into the 7 constructs. Next, the items were checked to either agree or disagree. The disagreements were resolved by further discussion. Second, the items were transformed where possible so that it is applicable to assess farming systems or technologies in general (eg. items 14, 25, 30, 123, 149, 164 etc.). For instance, item 14 “PA may not improve the efficiency of agricultural management.” was changed to “The system may not improve the efficiency of management.”, item 25 “I accept the consolidation of small farms into cooperatives to adopt PA.” was changed to “I accept the consolidation of my initiative into cooperatives to adopt the system.”, item 30 “I have the necessary knowledge to use PA technologies.” was changed to “I have the necessary knowledge to use the system.”, item 149 “Extension experts who have been providing consultancy in IPM&NM for a long time feel threatened by IPM&NM-DSS.” was changed to “The experts who have been providing consultancy for a long time feel threatened by the system.”. Third, RJT and RW removed items for which this was not possible (items 1-5, 17-19, 23, 39, 47, 50, 51, 62, 65-69, 86, 91, 93, 97-102, 110, 114, 126-129, 131, 135, 136, 139, 140, 148, 152, 154, 167-170, 176, 183 and 194). This reduced the list to 57 items.

Furthermore, we had a discussion to further cross-check the items. Due to low or no relevance in the actual farming domain, we removed some more items (item [20: 193, 81, 21], 25, 55, [79: 77, 83, 118, 107, 108], [106: 174], [122: 119, 120, 121], [182: 70, 71]). The sub-items were the ‘derived from’ items for the main item. Next, we transformed the language of some items to make them easy to understand (eg. item 96 “I enjoy using the system with an attractive user interface.” was changed into “I enjoy using the system with attractive graphics.”, item 143 “We possess those characteristics which a user of the system is supposed to have.” was changed to “This system was made for farmers like me.”, item 13 “The system may increase the costs and may not improve the revenue.” as changed to “The system may not show a positive cost to benefit ratio.” etc. Then we combined a few more similar items (item 105 with 155, 7 with 34, and 146 with 145). This reduced the list to 47 items used for the initial scale development as shown in Table 5, which also shows which original items these were derived from. Finally, we had a discussion to cross-check and correct the items for grammatical mistakes.

The developed scale of 47 shortlisted items needs to be validated with a user study to be used as a standardised scale in smart agriculture. The data from the study of the standardised scale will be analysed using Exploratory Factor Analysis (Principal Component Analysis -PCA- extraction) to cross-check if the items belong to a specified construct and if there are correlations. For the factor analysis (PCA), all the 5-point scale items will be considered as ordinal measures. This will be then followed by the Confirmatory Factor Analysis to determine the validity of the scale, and to confirm the constructs/factors. This will help in the cross-validation of the items across the scale and in the reduction of the total items. The validated Technology Acceptance Scale can be used to measure acceptance for an existing product. It can be also used in the co-creation or development of new tools and products in smart agriculture.

Reflection

The developed shortlisted scale of 47 items can be used to evaluate technology acceptance in crop farming. The scale was developed based on the literature screening on crop-based agriculture that focused specifically on intelligent technologies such as decision support systems without considering acceptance of broader technology in agriculture. We acknowledge the possibility of subjective judgement in shortlisting articles and the merging of the scale items. Additionally, the mapping of the shortlisted articles to the Theoretical Model of Acceptance is a subjective representation from the literature analysis. Therefore, these limitations may not necessarily transfer to the developed scale as it is a combination of all the collected scale items. However, this can be confirmed only after the data collection process when the scale will be implemented and validated to produce a shortlisted validated version for future use. The formulation of such a scale would contribute to closing the gap that exists in the present non-validated methods of technology acceptance evaluations in this domain. Future work across the community on a validated scale would allow us to frame research questions within technology acceptance constructs, allowing for the validation of a standard scale in crop farming and possibly the wider agricultural community.

Appendix