1 Introduction

Automation of repetitive physical work tasks is a long-running global trend. The number of industrial robot installations is estimated at 570,000 global units as of 2022. Between 2015 and 2021, installation of robots has more than doubled (IFR 2022). The food processing industry follows this trend (Romanov et al. 2022). The year 2020 was the first time in history where the non-automotive robot commissions surpassed those of the automotive industry, with a substantial increase in orders for food and consumer manufacturers (MarketScale 2021). There are several reasons for this.

One is the desire to reduce operating costs and increase productivity (Badri et al. 2018; Botti et al. 2015). Another is that the demand for food is expected to increase with the ongoing increase in the global population (van Dijk et al. 2021). However, because of the labour-intensive nature of food processing operations (OECD 2020), labour shortage is a concerning risk for food production, leading to increased interest in automating such processes.

A third reason is the public perception of the sector. The recent COVID-19 pandemic struck the meat sector workers heavily in several European countries compared to other sectors (Durand-Moreau et al. 2020; Schmidt 2020; Waltenburg et al. 2021). This brought to the public’s attention several social challenges within the meat sector, such as “[e]xtremely long working hours, piecework in confined spaces, lack of breaks and unhygienic collective accommodation at inflated rents” (Schmidt 2020). In this context, unions warned about exploitation of workers in some parts of the meat industry due to the use of subcontracted workers, kept on lower pay and more challenging working conditions than regular workers (McSweeney and Young 2021).

The increased use of robots, automation, and artificial intelligence (AI) in the food industry is raising questions about subsequent societal impacts. On the one side, automation and robotics in general are seen as a threat: 45–60% of all workers in Europe could see themselves replaced by automation before 2030 and about half of current jobs globally could in theory be automated (European Commission 2022). On the other side, automation and robotics can also be seen as an opportunity because workers threatened by automation can potentially find similar or less physically demanding work if appropriate training were provided. In Europe, for example, meat processing plants are increasing the rate of automation for some work tasks, sometimes with the motivation to make the work of the meat processors less physically demanding (Hansen 2018; Romanov et al. 2022). Pig slaughtering, cutting and deboning are traditionally very labour-intensive and require hard and repetitive work (Hinrichsen 2010). In this context, the introduction of automation can be seen as a positive opportunity for improving the working conditions. However, automation and robotics are often seen as suitable only for large-scale facilities with high running costs, while it is held to be less feasible for small-scale producers with lower productivity.

This article is a study of an innovative Autonomous Robotic System (from now on called ARS) for pork processing plants which has been developed to handle slaughtering, deboning, and cutting processes. The projected ARS, which is currently at the pilot phase, has highly autonomous features and uses core robotic technologies based on AI and cognition. It contains robots that lift, hold, cut, and debone the carcass. The development of the ARS could be attractive to smaller producers who cannot invest a large amount of money on automation but could also be scalable to larger producers when deployed in parallel (de Medeiros Esper et al. 2021). The ARS is not in physical contact with the operator and safety cameras are installed on the factory floor for shutting down the system if there is a risk of accidents. Further processing of the meat cutting and deboning is still mostly manual processes, relying on human operators, since these tasks require a higher level of dexterity and sense (de Medeiros Esper et al. 2021).

In the phase of transitioning from manual labour to automation, it is important to assess the potential social consequences that such a system can have on the society and how to develop it in a socially sustainable way.

Life Cycle Thinking (LCT) aims to transcend “the traditional focus on production site and manufacturing processes to include environmental, social and economic impacts of a product”, preferably across the entire product life cycle (Life Cycle Initiative 2023). Life Cycle Thinking “is a systemic and comprehensive concept considered pivotal to provide support in better integrating sustainability into policy making” (Sala et al. 2021). Life Cycle Assessment (LCA) is one of several tools developed to support LCT (ibid.), to give comprehensive quantitative information on the environmental performance of products and services, and to assess and support sustainable production and consumption (Guinée et al. 2011; Sonnemann et al. 2018). LCA has traditionally been applied to environmental impacts (Finnveden et al. 2009), but over the last couple of decades, Social Life Cycle Assessment (S-LCA) has also become more common (Huertas-Valdivia et al. 2020).

Social LCA may be called a highly fragmented research field where many challenges regarding methods, framework, paradigms, and indicators are still unsolved (Arcese et al. 2018; Huertas-Valdivia et al. 2020). Huertas-Valdivia et al. (2020) pointed out that even if S-LCA has been applied to different sectors (agriculture, bioenergy, transport, water management, chemical products, electronics, etc.), the geographical context cover mainly non-European countries.

To our knowledge, there are some few studies trying to characterise the social aspects of pork production such as the studies of Lagarde and Macombe (2013), Neugebauer et al. (2014) and Petit et al. (2018). Lagarde and Macombe (2013) highlighted that the changes in one part of the supply chain (in this case due to the introduction of a new pig plan farm) will affect the whole pig meat supply chain, especially regarding the number of jobs created/destroyed. Petit et al. (2018) proposed a list of indicators to match the sustainability requirements in the whole value chain of pig meat, including worker welfare and consumer information about the slaughterhouse. Neugebauer et al. (2014) examines the social performance of the pork production value chain, focusing on the stakeholder categories Workers, Local communities, Consumers, and Animals.

Valente et al. (2020) conducted a Life Cycle Sustainability Assessment (LCSA) on a semi-automated pig slaughter system, utilizing the principle of co-operative robotics, where robots performed relatively simple, repetitive and labour-intensive tasks (e.g., lifting, holding, stretching), while a human operator performed more complex functions (e.g., detailed cutting). The geographical scope was Norway, where there is still little automation compared with large slaughterhouses abroad (de Medeiros Esper et al. 2021; Mason 2022). This LCSA study included a S-LCA part that identified social risks mainly linked to the social subcategories ‘health and safety’, ‘equal opportunities and discrimination’ and ‘migration’ in the conventional slaughter and cutting process, with some degree of improvement in the semi-automated system.

Due to the increasing automation in food processing, it is important to understand the implications of automation so that issues relating to social performance and impacts may be addressed early in the process of developing new food automation solutions and technology, and not only after. Apart from the study by Valente et al. (2020), the authors have not been able to identify other studies where S-LCA is applied to the topic of automation and food processing.

This paper contributes to the state-of-the-art by assessing the social aspects of a future fully automated system for pig slaughtering. The aim of this study is to assess the social performances and impacts before the introduction of the ARS (pre-ARS or conventional line) and after its introduction (post-ARS scenario) at the slaughterhouse, focusing on the stakeholder categories assessed to be of highest relevance. We discuss the ARS in a general European context, with some specific Norwegian data when assessing social impacts.

The novelty of this paper is that it is an assessment of social impacts at an early stage of food processing technological development, assessing and comparing the present situation with a scenario, applying a novel combination of methods for acquiring data for such an anticipatory assessment. This contributes to methodological development in S-LCA, to the pool of S-LCA knowledge of food production development in Europe and to the responsible development of autonomous robotic systems in workplaces in this sector.

In the following sections, we will first provide a detailed account of the methods of the study, then we will present the results. In the following discussion section, we will interpret and discuss the results considering relevant literature, present limitations of the study and outline implications and further research needs. We will end with a brief conclusion.

2 Methods

S-LCA is a social impact assessment method for assessing the social and socio-economic impacts all along products and services life cycles (UNEP 2020). So far, it is not an ISO standard methodology, even if guidelines and handbooks exist and a process of standardization is ongoing (ISO/DIS 2023). The study follows the updated version of the Guidelines for Social Life Cycle Assessment of Products and Organizations 2020 of the United Nations (UNEP 2020). The conventional four phases of LCA, as described in the ISO 14040 standards (ISO 2006a, b) on environmental life cycle assessment, are followed (i.e., goal and scope, life cycle inventory, life cycle impact assessment, and interpretation). In S-LCA, there are two types of impact assessment approaches: the reference scale approach (known as type I) and the impact pathway approach (known as type II) (UNEP 2020). This study uses type I since it aims to describe the social performance of the pre and post ARS.

Figure 1 illustrates how the four phases were applied in this study. Each phase - (1) goal and scope: setting the goal (comparison of pre and post ARS) and defining its scope (object of study), (2) inventory: data collection strategy (collecting primary and secondary data by several techniques), (3) impact assessment (attributing inventory results to subcategories by a reference scale approach), and (4) interpretation of results (discussions of the results, conclusions and uncertainty analysis) - is described in detail in the respective Sections 2.1, 2.2, 2.3 and 2.4.

Fig. 1
figure 1

Graphical representation of the four phases of S-LCA. Pre-ARS data sources in yellow and post-ARS data sources in orange

Full size image

2.1 Goal and scope

The goal of the study is to assess the social impacts of the processing activities involved in the slaughter stage of the pork value chain (i.e., the result of the impact assessment step) and compare the social performances of introducing ARS at a meat processing plant (post-ARS) versus the current situation (pre-ARS), using a scoring system. The geographical scope of the study is Europe. In some cases, Norwegian data is used for presenting the results since the system has so far been located in Norway.

The timeframe of the study was from the year 2020 to 2022. The study is a gate-to-gate assessment since it focuses only on the core processes of the processing stage, i.e., at the slaughterhouse. The life cycle inventory (LCI) includes the inventory of the physical flows of the study system (the conventional line-based production system) normalized per functional unit, which was defined as 1000 kg of pork carcass at the slaughterhouse. The reference flow is therefore defined per 1000 kg pork carcass at the slaughterhouse in one year.

The system boundary of the study is presented in Fig. 2. The study focuses mainly on the activities at the slaughterhouse aiming to produce a half carcass (primary processing) in preparation to the primal cuttings and deboning (secondary processing). The primary and secondary steps may take place in either two separate zones of one facility or at different production plants. In a traditional slaughtering line, the primary and secondary processing is organized as follows: “Sequential steps, using overhead or belt driven conveyors to present the raw material to machines or human operators. In secondary processing, it is typical to see butchers stood alongside a conveyor, performing necessary operations on meat pieces. Configuration can vary from processor to processor, depending on requirements, volume of production and supplier of equipment. Usually, this type of line in the red meat industry is often referred to as a “pace line” and requires each of the operators involved in the process to work at a constant (and usually relatively high) speed” (Romanov et al. 2022).

Fig. 2
figure 2

System boundary. Source: Marel. Adapted by NORSUS

Full size image

In the S-LCA study, this delimitation is not straightforward. Often, social aspects are more connected to organizational behaviour than to a specific process, as also previously confirmed in Valente et al. (2020), which is why the system boundary also includes further processing. This study applied a cut-off criterion based on social significance in a qualitative way. We only include the processes which have the highest potential of being affected by the social changes introduced by the ARS as suggested by Dubois-Iorgulescu et al. (2018).

Animal production and therefore the live animal handling process, including animal welfare concerns, is outside the system boundary since the operation such as stunning and killing of the animals are conducted before the carcass enters the slaughter system (Alvseike et al. 2018). Packing and labelling are also outside the scope of the study since the ARS application will not affect these stages.

2.1.1 Social risks in the European pork processing sector

To identify and prioritize the stakeholder categories and subcategories, it is fundamental to understand the context of the study, i.e., the background situation for identifying the most relevant social hotspots (the activity in the life cycle where a social risk is likely to occur) by a sectoral risk analysis. The European Economic and Social Committee (Schmidt 2020) states that the working conditions in some meat facilities may be unacceptable, with the term “modern slavery” being used. Precarious contracts, low wages, extremely long working hours and a lack of breaks are widespread. In general, the meat sector in Europe has been accused of being characterised by a large extent of subcontracting, and the meat sector has been identified by the International Labour Organization as a sector in which workers face particular risk because of non-standard forms of employment and a high share of migrant or cross-border workers, who are often, allegedly, working under conditions with signs of exploitation, social dumping and unfair competition (ILO 2015). Meat processing is a low-wage industry which increasingly employs workers on temporary contracts hired by external recruitment agencies. Such sub-contracts typically mean that the liability for the workforce, its payment and working conditions is passed on entirely to the subcontractor. The European Federation of Food Agriculture and Tourism Trade Unions reports cases of subcontracted workers generally working 48–65 h per week and the working day can be up to 16 h, 6 days per week (EFFAT 2020). In comparison, workers employed directly by the meat companies generally work between 40 to 48 h per week. The report states that there is a general under-reporting of working hours in the sector, as handwritten time recording is common. In some European countries, overtime is often unpaid, especially in areas with high unemployment and illegal practices are widespread regarding working hours. Migrant workers are extensively used in the European meat industry. This can potentially impact both the local communities that are abandoned, and the local communities that attract migration. The gender wage gap is also a challenge in the meat industry (Schmidt 2020).

2.1.2 Stakeholder categories, subcategories, and indicators identification and prioritization

A S‐LCA study should always define which stakeholders are affected by the impacts. The scope of the study includes two stakeholder categories:

  • Workers

  • Local communities

These two stakeholder categories were selected and prioritized since they have a direct social relationship with the technology and are potentially impacted by changes due to the introduction of the ARS. Other stakeholder categories such as society, value chain actors, consumers, and children were excluded from the assessment since they could only be indirectly affected by these changes. The Guidelines for Social Life Cycle Assessment of Products and Organizations of UNEP (2020) recommend a set of impact subcategories (hereafter: subcategories) for each of these stakeholder categories. A subcategory is an impact category that is assigned to a stakeholder group, for example Working hours for Workers (UNEP 2020). Some of the subcategories initially appeared to have limited relevance to this assessment. Therefore, decision criteria following the general approach observed by Johnsen and Løkke (2013) are used to include or exclude impact subcategories according to the criteria: S-LCA database data availability, direct social significance to the industry and the meat sector, and relevance to the technology in the study. Only the subcategories included in the Methodological sheets for Subcategories in Social Life Cycle Assessment (S-LCA) (UNEP 2021) supplementing the Guidelines for Social Life Cycle Assessment of Products and Organizations published by UNEP (2020), were scrutinized using the decision criteria. See the Supplementary Information (SI) (Table S1) in the excel sheet named “decision matrix” for an overview of subcategories and a matrix summarising the initial decision of inclusion or exclusion in the scope of the study. Figure 3 illustrates the final selection of the stakeholder categories and subcategories included in the study. A short description of what it is evaluated in each subcategory is presented below the figure.

Fig. 3
figure 3

Stakeholder categories and social subcategories prioritized in the assessment

Full size image

For workers:

  • Freedom of association and collective bargaining: Whether the workers are free to form and join association(s) of their choice even when it could damage the economic interest of the organization, whether the workers have the right to organize unions, to engage in collective bargaining and to strike.

  • Health and safety: Incidents and injuries with and without sick leave, systems for reporting accidents, protective equipment, the extent to which workers are subject to noise, as well as the intensity of the work.

  • Fair salary: Practices concerning wages comply with established standards and if the wage provided is meeting legal requirements, whether it is above, at or below industry average and whether it can be considered a living wage (enough for supporting family need).

  • Working hours: Number of hours really worked is in accordance with the collective bargaining agreement and when overtime occurs, compensation in terms of money or free time is planned and provided to workers. Average number of hours of knife per working day, number of tasks, working breaks and type of rotation scheme are also other aspects included in this subcategory for evaluating the job tasks and repetitiveness.

  • Equal opportunity and discrimination: Management practices and the presence of non-discrimination actions in the opportunities offered to the workers by the organizations and in the working conditions. Gender and age equality is assessed in addition to gender wage gap.

  • Employment relationship: Worker contracts, e.g., whether contracts are used, whether subcontractors are used, and the degree of protection this provides for the worker. There is assumed to be some overlap between this category and the subcategory ‘Freedom of association and collective bargaining’ since certain worker rights are required by law regardless of the use of contracts.

For local communities:

  • Access to immaterial resources: The extent to which organizations respect, work to protect, to provide or to improve access to immaterial resources, especially education and trainings.

  • Migration and delocalisation: The extent of the migration within communities and whether populations are treated adequately. Temporary labour immigration is also considered in the assessment. Opportunities for communication and education between migrant workers and permanent residents to minimize risks of social dumping, and how well workers are integrated with permanent residents.

  • Local employment: The job impact, the local workforce, the turnover rate and all the aspects linked to providing job opportunities to community members.

Several indicators are selected for each subcategory, covering the most relevant social aspects for each subcategory. The selection of these indicators also follows the decision matrix approach of Johnsen and Løkke (2013) according to the following criteria:

  • Direct relevance to baseline case (pre-ARS)

  • Indicator likely to change with the introduction of the ARS (post-ARS)

  • Assumed data availability

  • Recommended by the methodological sheets for subcategories of UNEP (2021)

Furthermore, the literature review carried out by Valente and Johnsen (2022) on the topic of S-LCA and working environment was also taken into consideration as basis of the selection.

The final list of the indicators is presented in the SI in the sheet named “reference scale LCIA approach”.

2.2 Life cycle inventory

The Life Cycle Inventory phase includes the data collection of primary and secondary data as input to the research study for all unit processes included in the system boundary. Several tools were used in the inventory phase such as a S-LCA database, desktop research, questionnaire to industrial reference group, survey, semi-structured interviews and focus group interviews. Each tool with the explanation for its choice is presented in detail below.

2.2.1 Database

Social databases are useful for finding out the current most salient risks at country and sector level for specific stakeholder categories and subcategories. Thus, such databases can be used for identifying the social risks of the current situation in the pre-ARS scenario.

In this case, the Product Social Impact Life Cycle Assessment database (PSILCA) version 3 implemented in the software package openLCA version 10 (Maister et al. 2020), was used for screening analysis to find the potential social risks associated with the current Norwegian pork production. The social risk is evaluated in a scale from very low risk to very high risk. The background data follows a hybrid LCA approach where physical flows are monetized (i.e., multiplied by product prices collected from the most recent European statistics) and used as inventory input to the PSILCA database. Physical flows cover pig carcass, energy consumption, water, waste & sewage services. External costs and secondary activities are excluded from the assessment.

2.2.2 Desktop research

The desktop research is often useful in an S-LCA study for defining the social background context of the study and finding relevant information that is feasible to apply to an S-LCA study. Traditional literature review using the conventional search engines such as ScienceDirect is not sufficient as a data source, because the S-LCA scientific studies often are limited and do not cover the whole spectrum of social challenges in a specific context. Thus, open-source publications such as technical and non-technical reports, newspapers, websites, and statistical databases such as Eurostat are useful additional sources of information.

Search strings included all combinations of “S-LCA” and “work environment”, “automation”, “robotics” as well as combinations with their synonyms and “slaughterhouse”, “meat”, “meat industry”, “pork”, etc. with the focus on recovering articles that provide an insight into the S-LCA of pork sector.

2.2.3 Industrial reference group questionnaire

Data collection in S-LCA differs from traditional data collection in E-LCA which describes mass flows and processes. Companies are often used to provide data on CO2 emissions for E-LCA, but they are not used to report social data for S-LCA. During the timeframe of the study, an industrial reference group was established to support and facilitate the social LCA work by exchanging knowledge and helping with specific data collection with the aim of evaluating the social impacts of a current industrial system. The group included six meat processors from four European countries interested in the technology developed from both a technical and societal point of view. A questionnaire was distributed to the industrial members of the group and discussed with the group.

Four meat processors filled out the questionnaire with data for the entire slaughterhouse and specifically for each part of the facility, i.e., the slaughtering, cutting and further processing when possible. The data collection scheme was formulated for collecting quantitative data (e.g., the amount of working hours, the numbers of accidents at the meat processing industry, etc.), qualitative data (e.g., description of the personal protective equipment in use, the type of contracts, etc.), or semi-quantitative data (e.g., the presence/absence of unions, the presence/absence of minimum wages, etc.). Sensitive information in oral, written, or electronic forms is managed confidentially.

2.2.4 Survey of the social challenges

The online tool SurveyXact was used for creating customized questionnaire-based surveys (https://www.surveyxact.com/). The survey was open to responses for 5 months in total; from August 2021 to December 31, 2021. The survey was openly accessible on the internet. To ensure the anonymity of the respondents, the organizer of the survey never had access to the respondents’ names or contact information, and the survey (with a URL) was promoted via the head of the industrial reference group who distributed an invitation to targeted people in the meat processing sector in the European Economic Area.

Thirty-six questions were formulated for evaluating social challenges in the transition towards automation and identifying the level of agreement with statements regarding the introduction of an autonomous robotics system in the meat industry. The questions were framed following the inputs from the industrial reference group and concern the social aspects prioritized in the study (see Section 2.1.2). In addition, two open questions and space for leaving general comments concerning the consequences of the automation process were also included. The number of respondents opening the survey was 79, but only 58% completed the survey. Most respondents had a job title connected to the research and development sector (65.2%), followed by manufacturing (19.6%), other job title (6.5%), engineering (4.3%) and automation specialist (4.3%). 50% of the respondents answered to work in a meat-producing company or in another part of the meat sector (e.g., for a meat industry association). Furthermore, 50% worked for a large company (more than 250 employees), 30% for a medium company (50–249 employees) and 20% for a small company (10–49 employees).

2.2.5 Semi-structured interview

Semi-structured interviews with a research director at a meat processing plant, representatives of a meat association and of the union of food, beverage and allied workers, and an expert butcher were conducted for complementing the more formal types of data. A pre-determined set of open questions related to social themes prioritized in Section 2.1.2 were formulated. The interviews were conceived as participatory conversations that involved an interactive dialogue between the researcher and another person, as described by Swain and King (2022).

2.2.6 Focus group interview

A focus group interview with stakeholders was conducted in the fall of 2022. The focus group methodology may be used for eliciting opinions on a common theme among a group of interested individuals, which can then build on each other to develop perspectives (Acocella 2012). The focus group consisted of six key stakeholders who represented small-scale producers, large-scale slaughter facilities, butchers, industry organisations, and the food industry training office in Norway.

The focus group session began with a live demonstration of the ARS with a plastic carcass and video demonstration of the ARS with a real carcass to ensure that the focus group participants understood how the ARS worked. They were encouraged to ask questions. The focus group questions centred on the following themes:

  • How stakeholders believed the ARS would work in practice

  • The advantages and disadvantages of the ARS

  • How the ARS might affect the working conditions of the workers, both physically and mentally

  • If and how the ARS might affect the educational level and training requirements for workers

  • How the ARS might affect the composition of workers at the slaughterhouse, especially in gender and age distribution

  • What slaughterhouse management must do to ensure the ARS is safe and efficient once it has been implemented.

2.2.7 Data quality and data type

The geographical representativeness of the dataset is considered as fair assuming that the dataset can be applied to areas with similar production conditions. The technological representativeness is very good since an identical technology can be applied in another slaughterhouse in other countries in Europe.

In the assessment of the ARS, the level of resolution of data varied from source to source. Primary data points were collected from the industrial reference group, followed up by informal interviews, and by focus group interview, and survey. Secondary data points (i.e., not directly collected at the meat processing plants in a specific production site) were collected by database and desktop research. A data quality assessment was performed for each data point:

  • For database data from PSILCA: Based on the rules of the pedigree matrix of (Weidema 1996) by five criteria: Reliability of the underlying data sources (R); Completeness conformance (C); Temporal conformance (T); Geographical conformance (G) and Further technical conformance (F) on a scale from 1 to 5 (from very good to very bad).

  • Questionnaire from industrial reference group: Data points were rated from low to good quality, based on data availability and the information received. Primary data for the whole sample are considered of good quality, secondary data are considered of medium quality, and data representing a small sample or descriptive information only are considered of low quality. Furthermore, a lack of available data is also stated.

  • For interviews: the data quality is assessed as good since the informal interviews provided primary data completing and adding useful information to social aspects.

  • For the survey: The data quality is defined as good, although since the survey was an open questionnaire, we cannot assess the representativeness of the respondents to a complete population. Thus, the survey can only be assumed to provide indications of topics and not to provide representative data.

  • For the focus group interview: The focus group includes an acceptable variety of stakeholders, participants were engaged in the demonstration and the following conversation, and therefore our qualitative assessment is that all stakeholder views were well represented and that all view and topics of importance were adequately discussed.

The different data sources in total seem to provide a relatively complete view of the current situation and the hypothetical situation with the robot, within the goal and scope defined in our study, and as such the overall data quality of the study is assessed by the research group to be adequate.

The data types are different for each data source:

  • Quantitative risk level for the PSILCA database

  • Quantitative, semi-quantitative and qualitative for questionnaire to industrial reference group

  • Qualitative and quantitative for the desktop research

  • 5 and 3-point Likert scales for the survey

  • Qualitative for the interviews and focus group interviews

2.3 Impact assessment method

The S-LCA uses a Reference Scale Approach (also known as Type I or Reference Scale S-LCIA) (UNEP 2020, p. 28). A reference scale approach is developed for both pre-ARS and post-ARS with the aim of scoring each indicator in each social subcategory. The reference scale is established by adapting the definitions presented in the methodological sheets of UNEP (2021), in the Subcategory Assessment Method of Ramirez et al. (2014), in the Fishery Performance Indicator (FPI) framework of Anderson et al. (2015) and in the Product Social Impact Assessment Report of Harmens et al. (2022). An example of the reference scale developed for the social subcategory ‘equal opportunities and discrimination’ is presented below for the indicator ‘gender rate’:

  •  +1: Genders are represented equally in staff composition.

  • 0: One gender is slightly overrepresented in staff composition.

  •  −1: One gender is very overrepresented in staff composition.

Each indicator is assessed against the reference scale and scored with linear scores, where each scale level goes from −1 to +1, where a high score is better than a low score. Each score at indicator level is aggregated in a final overall score at subcategory level, without weighting.

Weighting was not conducted as this practice is controversial in the LCA context, due to the lack of comparability of indicator results (Huppes and van Oers 2011). Moreover, criteria for weighting would involve trade-offs between aspects that are difficult to approach in a robust scientific manner (Schmidt and Sullivan 2002; Johnsen and Løkke 2013).

In the pre-ARS, the final scores per subcategory helps to describe the current state of the pork production system, with a focus on its social performance. In the post-ARS, the final score defines if the social performance will be improved or not by the introduction of ARS compared to the same reference. The spreadsheet in SI presents the reference scale for each indicator per social subcategories with a score and note on the scoring assessment (“Reference scale LCIA approach” sheet).

2.4 Interpretation of the results

In this phase, the results from the LCIA phase are discussed and conclusions are formulated. The confidence in the result is assessed by an uncertainty analysis. The uncertainty score was adapted (from a fisheries-specific context to a non-specific context) from the FPI framework (Anderson et al. 2015) in a scale from A to C (from high confidence to low confidence). Using an uncertainty score enables reflection on the quality of each assessment (ibid.), and thus even in high uncertainty all indicators can be scored and be transparent about the confidence in each score. The definition of each uncertainty score is as follows (ibid.)Footnote 1:

  1. (A)

    We are highly confident (95%) that the given score is correct. Confidence can come from familiarity with the case study, the reliability of another expert source, a calculation based on reliable data, or large ranges of the underlying indicator for the given score that make another score highly unlikely. Note that it is confidence in the score that matters, and thus wide ranges for the underlying indicator associated with a score can support “A” quality, even in the case when information about the precise level of the underlying indicator is poor.

  2. (B)

    We feel that the given score is more likely than others.

  3. (C)

    We are making an educated guess based on best available information.

The results of the uncertainty analysis are shown as score in the SI spreadsheet in the sheet named “Reference scale LCIA approach”, and briefly discussed in the paper.

3 Results

Figure 4 illustrates the overall results of the social performance of the system, as output of the social impact assessment in a spider diagram (see also Fig. S1). The impact is expressed as a score, in a spider diagram at subcategory level for the two stakeholder categories: Workers (green) and Local communities (blue). Detailed results for each data sources are shown in the SI in the sheet “Reference scale LCIA approach”. Results from the database and the survey to even more level of details are also presented in SI, in the sheets named “Database results” (Table S2) and “Survey results” (Fig. S2). The results show that there will be an improvement in all social subcategories when moving from the pre-ARS to the post-ARS scenario. The subcategory Health and safety is most improved (from a score of −0.5 to +1.0, an improvement of 1.5), followed by Access to immaterial resources (from a score of −0.3 to +1.0, an improvement of 1.3). Working hours, Freedom of association and collective bargaining, Fair salary improved their score by 1.0 (from 0.0 to +1.0), while Equal opportunities and discrimination, Employment relationship and Migration and delocalisation only improved 0.5.

Fig. 4
figure 4

Overall results of the pre-ARS and post-ARS scenarios in a spider diagram. The results have been summarised into subcategories and stakeholder categories Workers (green) and Local community (light blue)

Full size image

Results are presented below at social subcategory as the outcome of scoring at indicator level. In some instances, Norway is mentioned as an example because the original technical concept was developed in this country. Since the level of details in the results per each indicator is very high, it is highly recommended to download the SI where each indicator is scored against the reference scale in the pre-ARS and post-ARS with explanation on how to reach the score, including data sources and literature references. Each score is also evaluated by an uncertainty analysis in a qualitative way as described in Section 2.4. Furthermore, the detailed results in the SI spreadsheet present: Overall results of the pre-ARS and post-ARS scenarios in a spider diagram. The results have been summarised into subcategories and stakeholder categories Workers (green) and Local community (light blue).

  • The spider diagram with the original link to the average score for the two systems (Fig. 1 “Overall results” sheet);

  • Specific results from the database PSILCA representing the risk level for the prioritized subcategories and corresponding indicators (Table 2 “Database results” sheet) and

  • A selection of results from the survey (Fig. 2 “Survey results” sheet).

3.1 Equal opportunities and discrimination (pre-ARS score −0.5; post-ARS: 0.0)

When investigating this subcategory, our data collection showed that for the pre-ARS t the proportion of women in the pork sector is strongly underrepresented and the age groups are skewed (few young workers and senior workers). It is difficult to recruit young people in the sector in Norway. There is a system in place for reporting complaints of discrimination, but it is difficult to assess the promotion of actions for increasing equal opportunities and diversity. The risk of gender wage gap is in general low in Norway. In the post-ARS, the way of working will change, becoming significantly less physically demanding and thus opening more opportunities for women, even if automation alone will not increase proportion of women as women are also underrepresented in automation-oriented studies and occupations. The result from the post-ARS studies indicates that there will more opportunities for young staff: while job opportunities for senior staff will be the same, automation might make it more interesting for young people and recruitment of young generation might become easier. Older generations might not be on board with this “revolution” in the job role, but many will have aged out of the job by the time the ARS is implemented. The system for reporting equality complaints is assumed to be the same or better if staff become better educated. ARS will not change gender wage gap.

3.2 Health and safety (pre-ARS: −0.5; post-ARS: 1.0)

For the pre-ARS, no fatal accidents have occurred in the abattoirs of the industrial reference group, but there are incidents with and without sick leave. According to a Norwegian informant, the number of accidents with sick leave is higher than the average in other sectors in Norway, especially in the cutting and deboning hall. Furthermore, the rate of absenteeism in the Norwegian food processing industry is higher than the Norwegian average. There are safety measures and protective equipment in place such as cutting gloves, kevlar gloves, tunic/armor, safety shoes, helmet/home cap, safety glasses, etc.

For the post-ARS, there is a consensus that health and safety of workers will be improved for many different indicators: better ergonomics (less heavy lifting), reducing the risk of injuries, less sick leave and improving worker safety. The ARS will reduce the number of accidents, the job will become more safe and less strenuous, will involve less heavy lifting and less physical labour, and will be more hygienic and more technology oriented. The work environment will improve in terms of noise and potentially humidity. Health and safety rules will become less rigid, e.g., helmets will no longer be required. In the post-ARS scenario, the sick leave rate and the injuries are expected to decrease, ergonomics and worker safety will be improved, even if the main driver for full scale automation and robotization of slaughterhouse is not directly related to employees’ safety and health.

3.3 Working hours (pre-ARS score: 0.0; post-ARS score: 1.0)

In the pre-ARS, the average number of working hours and overtime follow the rules of the national collective agreement. The number of hours spent in the same task is high and so the tasks are repetitive and only physical. In the post-ARS more flexibility in working hours and more part-time work is expected. There will still be two shifts and the same skillset on all shifts. There are no changes in the handling of overtime. It might lead to less rotation because shifts won’t be divided by activity, it will be continuous. The slaughter and cutter will be same person, which may make the meat processing stage less monotonous. People will work more in teams, i.e., butchers, technicians, engineers, which means that the work will become more social and the tasks will be more intellectual (working with the robot, problem solving and less physical (the robot does the heavy lifting).

3.4 Freedom of association and collective bargaining (pre-ARS score: 0.0; post-ARS score 1.0)

In the pre-ARS, the trade unionism is very high in the food processing industry, however there is also high use of agency workers which are likely not unionised. Agency workers, in particular migrant workers, might have lower degree of affiliation with unions. Sectoral collective bargaining and right to strike are in place and there is no evidence of the risk of fragmentation in the collective agreement,

In the post-ARS, there will be more skilled labour, which is more likely to be recruited locally (as opposed to agency workers) and therefore more likely to be unionised. Therefore, we estimate that trade unionism will be higher in the post-ARS scenario. Worker representatives need to be involved at an early stage of the automation and robotization process.

3.5 Fair salary (pre-ARS score: 0.0; post-ARS score 1.0)

In the pre-ARS, there is collective bargaining agreement for salary in place for the sector and the meat cutters have a fair wage. However, even if the meat processors follow the collective agreement regulating the sectoral wage for the meat sector there is still a theoretical potential risk that salaries for butchers are too low to cover the necessary living costs of an individual or family (living wage).

In the post-ARS, the salary is likely to increase slightly as ARS cannot be operated without skilled labour. There is less consensus in the salary improvement. However, few concrete data points were available in assessing this subcategory due to sensitive information in both cases.

3.6 Employment relationship (pre-ARS score: 0.0; post-ARS score: 0.5)

In the pre-ARS, most of the contracts are permanent. Most of the short-term contracts are in the further processing division(s), because of the seasonal variation (in the slaughter season/before Christmas). There are more long-term contracts in cutting and deboning since these operations are stable throughout the year. The meat processing industry has bigger seasonal fluctuations of employment throughout the year compared to dairy processing industry and there is a significant share of seasonal agency workers.

In the post-ARS, there is less consensus on whether the introduction of ARS will improve the working contract, even if the working conditions are generally expected to get better. The routine task jobs with regular contracts will be the same as before. With fewer agency workers and more permanent employees, there will be less use of temporary contracts and the risk of contract violations will be reduced, even if the phenomenon of agency workers is unlikely to be eliminated completely.

3.7 Access to immaterial resources (pre-ARS score: −0.3; post-ARS score 1.0)

In the pre-ARS, the proportion of professional educational certificates is quite high, over 70–80% in average, but few employees have other types or levels of education (bachelor’s or master’s degree). Training courses such as national language and occupational health and safety courses are offered to the workers.

The post-ARS will probably not be run by engineers, even if some engineers will still be necessary, but by people specialised in automation with professional qualification. ARS will most likely be operated by teams of engineers, automaticians and traditional butchers. Working in teams may be seen as a better solution than retraining staff. By the time the ARS reaches the market, a new generation of workers may be in place who will be ready for the technology. Since people will work more in teams there will be a need for cross-cutting competencies and training programmes including upskilling, since it will be difficult to work in ARS without necessary skills. There must be technicians who can fix problems when they arise to avoid down-time. Staff will have to be trained to handle the robots and to work in a team. Local communities will potentially gain a more qualified work force. Low qualified personnel are likely to lose the job, while personnel with higher educational level will have more job opportunities. Management will have to ensure staff have the right competences to operate the ARS.

3.8 Migration and delocalisation (pre-ARS score: −0.5; post-ARS score: 0.0)

In the pre-ARS, the average number of migrant workers is higher than the national average. According to the interview, in this sector, almost half part of the workers come from abroad and especially from Eastern Europe. Migrant workers are often employed in part time and temporary positions and are strongly overrepresented in manual jobs without educational requirements. This can lead to cases of social dumping. Migrants will potentially have a more transient job relationship and a lower degree of social cohesion, as well as less awareness about rights, conventions and existing institutions in the country and sector where they work. Local workers, on the other hand, will potentially possess more awareness about and therefore access to collective associations that can be used to forward their rights and interests. At the same time, an increased migration rate can potentially undermine the bargaining power of local workers. In the post-ARS, there might be fewer job opportunities for migrant workers or lower skilled staff due to increased requirement for high skilled staff. This is not necessarily a negative development because companies already struggle to recruit workers, both local and migrant/agency workers. The ARS will not change the nature of some jobs, so there might still be an opportunity for migrant/agency workers there.

3.9 Local employment (pre-ARS score: −0.2; post-ARS score: 0.7)

In the pre-ARS, most of the workers (excluding the agency workers) are employed and live in the surrounding local communities, even if the percentage of workers hired locally is lower in the cutting and deboning. Labour shortage is a challenge in the current meat processing industry and there is a need for recruitment agencies. The post-ARS may improve the economic development of the local community as it would allow smaller slaughterhouse units to be built with more geographical spread. This means that ARS will allow for decentralisation in the meat industry and move the value chain back into the districts where it might be easier to recruit local people. The staff turnover can be hypothesised to be lower in smaller, more geographically spread slaughterhouses than in centralised pre-ARS slaughterhouses.

4 Discussion

The aim of this study was to assess the social performance and impacts of an autonomous robotic system (ARS) for meat processing in a pre- and post-implementation scenario. In this section, we discuss the results of the study considering this aim and existing literature, the strengths and limitations of the study, and, finally, implications and further research.

4.1 Interpretation of the results

The results are discussed at social subcategory level for the stakeholder categories Workers and Local community only focusing on the social subcategories showing the largest improvement in the social performance (the difference in the final scores between the pre-ARS and post-ARS was higher than 1) and the ones with the smallest improvement (the difference was lower than 1).

4.1.1 Workers — Health and safety (large improvement)

Having robots perform arduous, repetitive motions and tasks instead of humans would improve the ergonomics in the meat industry (Botti et al. 2015). However, scientific literature dealing with health and safety issues in automated or robotized slaughterhouses remain scarce. This is probably because the main driver for full scale automation of slaughterhouses is not primarily related to employees’ safety and health, but instead increased productivity and lower operating costs (Badri et al. 2018; Botti et al. 2015). Neugebauer et al. (2014) found that the biggest burden on workers at slaughterhouses was the pressure to perform a high number of slaughters per hour, thus being exposed to a high psychological stress. This is an impact that is relevant to think of in the context of automation.

4.1.2 Workers — Equal opportunities (small improvement)

We found that automation alone will not solve the gender balance issue in slaughterhouses. According to Madgavkar et al. (2019) women account for less than 25% of machine operators and craft workers. The same authors also found the effect of automation to have a similar impact on both genders, displacing or gaining jobs (about 20% of both genders affected). However, women could be more prone to having their jobs automated than men and would therefore need to update their skills. Thus, promoting gender equality in such a way as to avoid recruiting a larger share of male workers will remain important (Prestrud and Valente 2022). Related to discrimination, Aksoy et al. (2021) found that a 10% increase in robotization in 20 European countries leads to a 1.8% average increase in the gender pay gap, especially in countries in which initial overall gender inequality was high. The Technology and Innovation Report of United Nations (UNCTAD 2021) identified a historical correlation between technological change and inequality.

4.1.3 Workers — Employment relationship (small improvement)

Our findings indicate that there is not a substantial improvement in the subcategory ‘Employment relationship’. The ARS might cause a larger number of freelancers in the job market, which would cause a decline in traditional labour rights (Hansen 2018). However, investigating the conditions for the agency workers group is challenging because of their “structural and spatial invisibility” (Lever and Milbourne 2017). For improving the welfare of workers, social aspects such as personal capacity development and the involvement of employees in company practices should be considered in addition to worker wellbeing. The introduction of a smart work solution where the ARS at the abattoir can be controlled remotely, e.g., from home or even from another country, can improve the working environment at the slaughterhouse, but can also be associated with more insecure working contracts, creating a precarious class of on-demand workers.

4.1.4 Local community — Access to immaterial resources (large improvement)

Especially in the starting phase of ARS, highly skilled workers are expected to manipulate, operate, and maintain this complex system. The skillset required is foreseen to be different from the one typically possessed by the actual work force having low level of education. This finding is also supported by literature (Rotz et al. 2019; Seaton 2022).

4.1.5 Local community — Migration and delocalization (small improvement)

Migrant workers, often as low educated workers, are at risk of losing their jobs. Decentralized production becomes possible with the ARS, indicating a domestic geographical movement of work which could strengthen certain local communities, who might gain more jobs performed by qualified personnel. Involuntary resettlement might be another issue connected to this domestic movement.

4.1.6 Local community — Local employment (small improvement)

Automation will result in job losses (Frey and Osborne 2017), but it seems like jobs are impacted by automation, but rarely disappear (Anzolin 2021). The automation often substitutes a specific task, but not a “whole job”. However, automation across industries has been shown to negatively affect employment (Anzolin 2021). A previous LCSA study of automation in pig production showed that a semi-automated system would create fewer jobs as the system was expected to be more efficient (Valente et al. 2020). This study concluded that a fully automated pork sector would likely see a reduction in the traditional job of butchers while there will be a need for other skills to operate and maintain automated systems. Therefore, it is unlikely that this new technology will lead to net positive job creation, i.e., the sum of new jobs created from the new technology minus the jobs lost in the conventional line technology. Bringing ARS to the slaughterhouse might also have side-effects on the job market in other industrial sectors, such as transport (to move larger amounts of produced meat and transport robots to site), energy (to operate plants and robots) and education (to train skilled workers if demand increases for both building, operating, and maintaining robots and to support continuous education of labour force), to mention a few examples.

4.2 Limitations of the study

This paper has taken a practical approach to overcoming some of the conventional challenges related to the application of S-LCA, such as difficulties in prioritizing social indicators and subcategories, difficulties with data collection (Sala et al. 2015), and a lack of a transparency (Huertas-Valdivia et al. 2020). In the application of the reference scale-based LCIA-approach to ARS, the choice of subcategories and indicators for the social subcategories based on a decision criterion following the approach of Johnsen and Løkke (2013), proved useful. The indicated confidence level in the scores and the data quality assessment for each data point give a more transparent picture of the results. Higher confidence level in the score (A), combined by a good quality level of the datasets, make the results more reliable.

Still, there are challenges regarding data availability and low data quality level for some indicators, especially related to geographical representativeness. Some of the background information from the Norwegian perspective are relevant for the discussion of this technology in general, considering that the technology can be applied to areas with similar production conditions. However, the ARS is supposed to solve challenges in a European perspective where there are potentially large variations in conditions.

Furthermore, it is challenging in S-LCA to evaluate the social impacts and performances of a technology not yet in place, mainly due to the lack of data for the future scenario (Lehmann et al. 2013). For access to more and better data, survey and focus group interviews showed their strengths in collecting knowledge and opinions of a wider range of respondents, especially about a future scenario. For example, the survey indicated that a large share of the respondents was positive towards the introduction of the ARS, even if unsure whether the introduction of ARS will improve the social, environmental, and economic development of the local community.

The pork processing sector is highly diversified, with huge differences in slaughtering and cutting methods and slaughterhouse sizes across the Europe: from small-scale slaughtering to large-scale industrial installations. The study has covered only some aspects of this variability, presenting the results for social impacts in a European context.

4.3 Implications of the study and further research

There are several practical implications following from this study, which were already addressed with stakeholders in the focus groups. If such a cell-based, automated system is to be implemented, it has implications for training, re-skilling, recruitment, organisational structure, and leadership (include building organisational cultures). Abattoir owners and leaders should engage with the unions and with representatives of local societies to introduce such technological changes in a way that avoids unnecessary conflicts.

The need for such measures can be established already from the current research, but there should be subsequent research following-up on this study as the technology is implemented. It follows from the current study’s insight that implementation of automation in the pork processing industry is not only a technological challenge. It also requires changes in organizational aspects with respect to “tasks, personal skills, management approaches and process control” (Hinrichsen 2010). Since the employees are supposed to work more independently and take more responsibility than in the traditional slaughter line, the employees will have more pressure on their work situation and the supervisors must learn to manage employees with these new expectations, and not just provide instructions about what to do and how to do it. An implication of this is that more social science research is needed accompanying future, potential implementation processes, preferably with a learning perspective embedded.

In addition to this organizational perspective, further research should also address how to include other social aspects such as food hygiene and consumer acceptance.

Finally, we have in this study identified issues with ethical implications. The potential transition to ARS will open up for questions about the future societal role of artificial intelligence systems, requiring further research into the ethical and legal aspects of AI in the workplace. An ethical question for further research is also related to animal welfare. Many studies focusing on pig farms and slaughtering conditions point out the importance of including the social category of animal welfare in S-LCA (Neugebauer et al. 2014; Scherer et al. 2018; Tallentire et al. 2019; Zira et al. 2020). Even if animal welfare was defined as out of scope in this paper, the authors see that improved animal welfare could result from reduced transportation, as the ARS could facilitate more distributed abattoirs. Further research should also address the social justice dimensions of market and profit concentration, which might widen the technological gap between developed countries and less-industrialised countries.

5 Conclusions

This article has presented a study of the critical aspects before and after the introduction of an autonomous robotic system under development for the meat processing industry. The results show that the benefits outweigh the risks in the prioritized social subcategories: Equal opportunities and discrimination, Health and safety, Working hours, Freedom of association and collective bargaining, Fair salary, and Employment relationship for the stakeholder category Workers. In the stakeholder category Local Community, the benefits outweigh the risks in the social subcategories Access to immaterial resources, Migration and delocalization, and Local employment. The largest social improvement when moving from the pre-ARS to the post-ARS was in the Health and safety and Access to immaterial resources subcategories, while the lowest improvement was in the subcategories Equal opportunities and discrimination, Employment relationship, Migration and delocalization, and Local employment. Since the ARS scenario is still under development, and due to the complexity involved in stakeholder categories, a certain level of uncertainty should be considered in the interpretation phase.

A mixed method approach consisting of quantitative, semi-quantitative and qualitative tools was used. Some methodological development was necessary for the purpose of this anticipatory study, which can prove useful for the S-LCA research community.

The use of autonomous robotic systems will require appropriate training, alignment with new education programs and industrial organisational changes. Furthermore, potential topics to include in future studies will be to assess the impact of such systems on food hygiene and animal welfare, consumer acceptance issues, and the ethical implications due to the use of AI in the workplace. Social justice dimensions should also be further explored.