Graduate Student Literature Review: Social and feeding behavior of group-housed dairy calves in automated milk feeding systems*

Automated milk feeders (AMF) allow farmers to raise calves in groups while generating individual records on milk consumption, drinking speed, and frequency of visits. Calves raised in groups benefit from social interaction, which facilitates learning and adapting to novelty. However, calves in large groups (>12 calves/ feeder) experience a higher risk of disease transmission and competition than those housed individually or in smaller groups. Therefore, if group size, grouping strategy, and disease detection are not optimal, the health and performance of calves can be compromised. The objectives of this narrative literature review, from publications available as of February 2023, are to: 1) describe the use of AMF in group housing systems for calves and the associated feeding behavior variables they automatically collect, 2) linking feeding behavior collected from AMF to disease risk in calves, 3) describe research on social behavior in AMF systems, and 4) introduce social networks as a promising tool for the study of social behavior and disease transmission in group-housed AMF-fed calves. Existing research suggests that feeding behavior measures from AMF can assist in detecting bovine respiratory disease (BRD) and enteric disease (ED), which are common causes of morbidity and mortality for pre-weaned dairy heifers. AMF records show reduced milk intake, drinking speed, or frequency of visits when calves are sick. However, there are discrepancies between published research about the sensitivity of feeding behavior measures as indicators of sickness, likely due to differences in feeding plans and disease detection protocols. Therefore, considering the influence of milk allowance, group density, and individual variation on the analysis of AMF data is essential to derive meaningful information used to inform management decisions. Research using dynamic social networks derived from precision data show potential for the use of social network analysis (SNA) to understand disease transmission and the effect of disease on social behavior of group-housed calves.


INTRODUCTION
The individual assessment of calves' health and welfare with precision technologies can improve the efficiency of dairy calf rearing.Automated milk feeders (AMF) continuously record the individual feeding behavior of calves raised in groups.These systems allow for different milk allocation settings (Borderas et al., 2009a;Jensen, 2004) and report milk consumption, drinking speed (volume/time), the number of rewarded (i.e., the calf enters the feeder station, and a portion of milk is available) and unrewarded visits (i.e., the calf enters the feeder station, but milk is not allocated).Because feeding behavior reflects an animal's motivation to eat, changes in that behavior can be used to identify sick animals (Knauer et al., 2017;Cantor and Costa, 2022; reviewed by Morrison et al., 2021).Sick calves consume less milk, perform fewer visits, and drink milk at a slower rate compared with healthy calves (Knauer et al., 2017;Cantor and Costa, 2022).However, the changes in feeding behavior that sick calves display can differ depending on milk allowance (Borderas et al., 2009b) and disease (Knauer et al., 2017).Calves also modify their feeding behavior when they experience competition, by increasing their drinking speed and frequency of visits (Jensen, 2004).The appropriate use of precision data from the feeder can make it possible to identify sick animals promptly with predictive models (Bowen et al., 2021).
The most common causes of morbidity and mortality among pre-weaned dairy heifers in the United States are bovine respiratory disease (BRD) and enteric disease (ED).Bovine respiratory disease is a disease complex caused by one or more pathogens including bacteria, respiratory viruses, and Mycoplasma spp.(van der Fels-Klerx et al., 2002).Enteric disease is a multifactorial disease involving viruses, bacteria, and protozoa, and non-infectious factors such as management and feeding with clinical symptoms including diarrhea (Cho and Yoon, 2014).In 2014, ED accounted for 56.4% of deaths and 16.0% of antibiotic treatments, and BRD accounted for 24.0% of deaths and 11.4% of antibiotic treatments in dairy heifers (USDA, 2018).Timely detection of BRD and ED is important for appropriate intervention (McGuirk, 2008); however, detection of BRD and ED on commercial farms often relies on visual scoring systems which can be challenging to implement in group housed animals.
The Wisconsin scoring system for BRD is based on 4 clinical signs: temperature, nasal discharge, cough and ocular discharge, and ear position (McGuirk, 2008).In addition to the clinical signs the California scoring system for BRD adds abnormal respiration and head tilt (Love et al., 2014).These detection methods may be hard to implement in group settings and have moderate sensitivity and specificity (Buczinski et al., 2015).Lung ultrasonography is an alternative for BRD detection with higher sensitivity and specificity than the Wisconsin scoring system (Buczinski et al., 2015) but requires handling of the animals and specialized equipment.Calves with ED have higher intestinal secretions and malabsorption of fluids (Berchtold, 2009).Fecal scoring relies on fecal consistency as indicator of the water content of feces (Renaud et al., 2020), which can be difficult to observe for individual calves in group pens.These challenges can result in a higher risk of failing to provide sick calves appropriate intervention or treatment, compromising their performance (McGuirk, 2008).
The objectives of this narrative literature review, from publications available as of February 2023, are to: 1) describe the use of AMF in group housing systems for calves and the associated feeding behavior variables they automatically collect, 2) describe research linking feeding behavior collected from AMF to disease risk in calves, 3) describe research on social behavior in AMF systems, and 4) introduce social networks as a promising tool for the study of social behavior and disease transmission in group-housed AMF-fed calves.Publications about feeding behavior in AMF-fed dairy calves indicating or describing the symptoms of ED and/or BRD were considered for review.In the case of social behavior, studies involving group housed dairy calves fed with AMF were prioritized, however, studies in which calves were manually fed or pair housed were also reviewed.

Group housing for calves and AMF-systems
Calves are often individually housed during the preweaning phase, which was traditionally thought to prevent disease transmission (Jorgensen et al., 2017) and to provide customized care and feeding (USDA, 2016).In 2014, individual housing of pre-weaned calves was still the most common practice among dairy farms in the United States (74.9%;USDA, 2016).Common housing systems include hutches, individual pens, and elevated stalls in which calves are restrained by fencing or chain and collar (reviewed by Endres and James, 2017).When housed individually, calves drink predefined milk volumes at fixed times from bottles, buckets, and, more recently, robotic arms (reviewed by Endres and James, 2017).Limited social interaction in these systems can compromise a calf's ability to cope with novelty and stressful events such as weaning (de Paula Vieira et al., 2010).Early contact with other individuals is essential for calves' cognitive (Gaillard et al., 2014) and social development (Bøe and Faerevik, 2003;Lindner et al., 2021).Calves that experience social contact early in life are less fearful, more exploratory, and can adapt to a novel feed (de Paula Vieira et al., 2012) or unfamiliar calves (Lindner et al., 2021).
In a survey, Canadian producers indicated consistency of feeding and more frequent feeding as factors that influenced them to switch from manual feeding to AMF systems (Medrano-Galarza et al., 2017).With AMF, producers can feed larger volumes of milk, which calves can consume in smaller meals, without additional labor (Sinnott et al., 2021).This equipment also allows for individual record-keeping and gradual weaning, as controlled by age or other factors such as starter intake (Benneton et al., 2019;Welk et al., 2022).However, housing calves in groups with AMF can lead to competition, increased disease transmission, and present challenges related to disease detection.Thus, the success of these systems depends, among other factors, on the optimal combination of grouping strategy, group size, feeding plan, and data management.

Group size in AMF-systems.
Group size plays a role in feeder availability and stocking density, which are related to competition (Jensen, 2004) and disease risk (Svensson and Liberg, 2006) of group-housed calves.In one experiment, the incidence of BRD was significantly higher in groups of 12-18 calves than in groups of 6-9 calves (Svensson and Liberg, 2006).However, group size did not impact the likelihood of diarrhea in this study (Svensson and Liberg, 2006).Calves are motivated to synchronize feeding (Miller-Cushon and DeVries, 2015), resulting in competition, especially under an elevated calf to feeder ratio (Jensen, 2004).Displacements and acts of aggression among calves while drinking milk were more frequent in groups of 24 calves in an AMF system with access to one feeding station than in groups of 12 (Jensen, 2004).Reducing the ratio of nipples to calves Montes and Boerman: Literature review: Social and feeding behavior of pre-weaned calves that were fed ad libitum milk in one bucket (4 nipples: 3 calves vs. 1 nipple: 3 calves) reduced the amount of milk calves consumed and the time they spent drinking (von Keyserlingk et al., 2004).However, it has been observed that pair housed calves adapted to moderate competition by increasing meal frequency to maintain milk intake (Miller-Cushon et al., 2014).Therefore, housing calves in groups of 12 or fewer calves per feeding station can reduce aggressive interactions and disease risk in AMF-systems.

Grouping strategy in AMF-systems.
Groups of calves in AMF systems can be classified as either stable or dynamic.In stable groups, all the calves enter the pen within a few days and move out together when the last calf is weaned.In dynamic groups, new calves enter the pen and calves leave the feeder continuously.Pre-weaned calves in stable groups had increased weight gain (60 g more/day) compared with calves in dynamic groups (Pedersen et al., 2009).Whereas calves in dynamic groups had 2 times more treatments for ED and BRD compared with calves in stable groups (Pedersen et al., 2009).Therefore, maintaining pre-weaned calves in a stable group can improve performance and reduce the risk of disease.
Age of introduction and housing before introduction to the AMF have been evaluated to determine their impacts on grouping strategies.The age of introduction to groups can impact the assistance calves need to learn how to use AMF and their risk of disease.Medrano-Galarza et al., (2018) reported that calves that were introduced into dynamic groups that were less than one d of age took longer to use the feeder independently and experienced a higher risk of severe diarrhea than calves introduced at 5 d of age.Similarly, Jensen (2007) reported that calves introduced at 6 d of age into a dynamic group required more guidance to use the feeder and consumed less milk than those introduced at 14 d of age.Housing calves of similar age in pairs before grouping did not reduce the assistance calves needed to learn how to use the AMF (Lidner et al., 2021).Research suggests that calves allocated to an AMF should be backgrounded until at least 14 d of age to reduce manual training time, and 5 d of age to minimize the risk of severe diarrhea bouts.
Feeding behavior variables collected in AMFsystems An advantage of housing calves in groups with AMF systems is the ability to automatically record feeding behavior.When calves are experiencing pain or sickness, they exhibit reduced appetite and motivation to eat (Ames, 1997).Multiple studies have shown that reduced drinking speed, milk intake, and visits to the feeder can indicate that a calf is sick (reviewed by Mor-rison et al., 2021).In addition to disease, the number of animals in the group and the milk allowance can affect feeding behavior (Rosenberger et al., 2017).Moreover, the analysis of within and between individual variation of AMF records showed that it is possible to identify behavioral patterns for feeding rate and frequency of visits (Carslake et al., 2022).Therefore, the feeding behavior patterns of sick calves must be observed in the context of group size, disease, and milk allowance (Conboy et al., 2021;Sutherland et al., 2018) and consider individual variation (Carslake et al., 2022).
Milk allocation Milk allowance and meal size are strong drivers of feeding behavior (Jensen, 2004).In the United States, on average, pre-weaned calves are fed 5.7 L/d of milk or milk replacer, commonly divided into 2 feedings (Urie et al., 2018).Producers can program AMF to deliver different combinations of daily allowance, the maximum volume of milk allocated per feeding bout, and the interval time between feeding bouts.Studies in AMF have shown that the number of visits, feeder occupancy, and feeding patterns depend on the amount of milk offered per day (Borderas et al., 2009a;de Paula Vieira et al., 2008;Jensen andHolm, 2003, Nielsen et al., 2008), the number of milk portions (Jensen, 2004;Nielsen et al., 2018), and milk flow (Nielsen et al., 2018).
Calves offered limited milk volumes (10% of body weight) keep visiting the feeder even when it does not allocate additional milk, resulting in unrewarded visits (de Paula Vieira et al., 2008).Increased unrewarded visits have been observed during the weaning phase when milk allowance decreases (Borderas et al., 2009a).A study evaluating 4 different milk allowances (6, 8, 10, and 12 L/d) reported that calves offered less milk, performed more unrewarded visits both before weaning and during weaning (Rosenberger et al., 20017).Similarly, calves offered a lower milk allowance performed more and longer unrewarded visits than calves offered a higher milk allowance (≤4.8 L/d vs. ≥ 7.2 L/d, Nielsen et al., 2008).In a different study, the combination of reduced milk flow rate (by partial blockage of the hose connected to the nipple) and low milk allowance (≤4.8 L/d) increased the duration of unrewarded visits (Jensen and Holm, 2003).In contrast, calves having ad libitum access to milk not only consumed twice as much milk as those under restricted plans, but they also had smaller, more frequent meals and rarely made unrewarded visits to the feeder (de Paula Vieira et al., 2008).Existing research suggests limit-fed calves (i.e., ≤ 6 L/d) perform unrewarded visits as consequence of hunger (de Paula Vieira et al., 2008;Jensen and Holm, 2003).
With AMF, farmers have flexibility in the development of milk-feeding plans.However, these settings Montes and Boerman: Literature review: Social and feeding behavior of pre-weaned calves can influence how frequently calves visit the feeder and the time spent at the feeder.For instance, calves in a restricted daily milk allowance (i.e., ≤ 6 L/d) perform frequent, long, unrewarded visits.Whereas under a higher milk allowance, rewarded visits are more frequent, and unrewarded visits are rare.Therefore, it is important to consider the milk allowance when evaluating feeding behavior patterns in AMF, especially when comparing the results across research conducted with different settings.

Feeding behavior and disease in AMF-systems
Changes in feeding behavior can complement the visual scoring systems for ED and BRD detection since feeding behavior patterns are affected by the immune response.When the cells of the innate immune system identify the presence of infectious agents or tissue damage, they release cytokines which bind to brain receptors inducing fever, anorexia, sleepiness, decreased activity, and depression (Tizard, 2008).As consequence of the reduced appetite and lethargy, sick calves are less motivated to consume milk and visit the feeder.Because these parameters are continuously monitored by AMF, research has focused on identifying associations between feeding behavior and health status relative to the time of diagnosis (reviewed by Morrison et al., 2021).Table 1 summarizes the results from research studies that evaluated the associations between AMFrecorded variables and health status of dairy calves.The integration of AMF data into machine learning models can help determine deviations from baseline feeding behavior that indicate sickness in individual calves (Bowen et al., 2021;Ghaffari et al., 2022).
Sick calves have reduced appetite and consequently consume less milk (Lowe et al., 2019).Higher milk consumption was associated with a lower risk of calves having BRD (Perttu et al., 2023).Sick calves consumed less milk compared with healthy calves when treated for BRD (Borderas et al., 2009b;Cantor and Costa, 2022;Swartz et al., 2017;Duthie et al., 2021) and ED (Borderas et al., 2009b;Knauer et al., 2017;Lowe et al., 2019;Sutherland et al., 2018).However, the sensitivity of feeding behavior as an indicator of pain (disbudding; Sutherland et al., 2018) and disease is related to milk allowance (Borderas et al., 2009b).In a study with 2 different milk allowances (4 L/d and 12 L/d), calves in the restricted group did not reduce their milk consumption when sick compared with the group with higher milk allowance (Borderas et al., 2009b).In contrast, a study in which calves had access to up to 10 L/d reported reduced milk consumption as a promising indicator of initiation of BRD (Cantor and Costa, 2022).Calves may only exhibit reduced milk consumption due to sickness when they are fed greater amounts of milk (i.e., > 6 L/d).
Specific feeding behaviors displayed by sick calves that consume less milk include fewer rewarded visits to the feeder and consume smaller portions on each visit.A reduction in the number of rewarded visits per day was associated with BRD (Duthie et al., 2021;Knauer et al., 2017) and ED (Knauer et al., 2017) treatments.However, sick calves can adjust by consuming smaller volumes of milk per visit (Conboy et al., 2021).For instance, Perttu et al., (2023) observed that the higher the number of rewarded visits, the higher the risk of calves having BRD.Their analysis also showed a complex relationship between calf health status, total milk intake per day, and number of rewarded visits (Perttu et al., 2023).Other studies found no association between the number of rewarded visits and BRD (Conboy et al., 2021) or ED (Conboy et al., 2022).
Unrewarded visits are important indicators of disease when calves have access to a limited volume of milk (Svensson and Jensen, 2007;Borderas et al., 2009b).When calves remain hungry, they visit the feeder even if they do not receive more milk (de Paula Vieira et al., 2008).However, when sick calves are less hungry and less likely to spend energy on non-vital activities such as additional visits to the AMF (Cantor and Costa, 2022).When fed restricted volumes of milk (≤10 L/d), calves treated for BRD (Cantor and Costa, 2022;Johnston et al., 2016;Knauer et al., 2017;Morrison et al., 2022) and ED (Knauer et al., 2017;Lowe et al., 2019;Morrison et al., 2022;Sutherland et al., 2018) performed fewer unrewarded visits than healthy calves.In contrast, when fed larger volumes of milk (≥16 L/d) calves treated for BRD did not perform fewer unrewarded visits than healthy calves (Duthie et al., 2021).
In the case of drinking speed, faster speed was associated with a lower risk for BRD (Perttu et al., 2023).Cantor and Costa (2022) and Knauer et al. (2017) observed differences in drinking speed between calves treated for BRD and healthy calves on the day of treatment.Morrison et al., (2022) reported reduced drinking speed in sick calves compared with healthy calves 4 d before BRD or ED treatment.Calves treated for ED also had slower drinking speed than healthy calves (Conboy et al., 2021;Knauer et al., 2017;Lowe et al., 2019).However, other studies found no association between BRD status and drinking speed (Duthie et al., 2021;Johnston et al., 2016;Svensson and Jensen, 2007).Since the hose diameter modifies milk flow (Jensen and Holm, 2003), which can influence drinking speed (Haley et al., 1988), variations in the equipment may explain differences between studies.
Differences among the studies on the association between feeding behavior variables with BRD or ED can 5.The detection protocol or scoring system described was created specifically for the study.Milk intake, drinking speed, unrewarded visits, and rewarded visits are in a daily basis.
Rewarded visits: the feeder allocates milk to the calf.Unrewarded visits: the feeder does not allocate milk to the calf, because it is not permitted according to the feeding plan.
BRD: bovine respiratory disease, ED: enteric disease, ILL unclear diagnosis, infection, or fever, and General makes no distinction on disease.
NA indicates that the study did not observe that behavior.Table 1 (Continued).Summary of published research on disease and feeding behavior of calves in automated milk feeder systems be attributed to the differences in sickness detection protocols (reviewed by Morrison et al., 2021).Research suggests that drinking speed can be a useful indicator of disease, and milk intake can be a good indicator when milk allowance is high (>6 L/d).Nevertheless, Conboy et al. (2021 and2022) showed that despite being associated with health status individually, drinking speed, milk consumption, and unrewarded visits had insufficient accuracy to serve as a diagnostic test.Combining feeding behavior and activity parameters can provide more robust indicators of disease (Lowe et al., 2021).

Disease detection.
Even though multiple studies have reported changes in feeding behavior patterns that are associated with BRD and ED, fewer studies have evaluated the ability of AMF records to predict disease in pre-weaned calves.Knauer et al., (2018) used standardized self-starting cumulative sum charts to predict ED and BRD based on feeding behavior variables from the AMF.A combination of unrewarded visits, milk intake, and drinking speed had 0.64 sensitivity and 0.37 specificity for detecting sick calves.Similarly, Conboy et al. (2022) found that optimal cut points on cumulative milk intake and unrewarded visits resulted in 0.69 sensitivity and 0.22 specificity to predict ED.These 2 studies found no advantage in using feeding behavior from the AMF for diagnosis but agree it can be a screening tool.
Research using supervised machine learning algorithms reported improved potential in AMF variables to predict disease in pre-weaned calves (Bowen et al., 2021;Cantor et al., 2022;Ghaffari et al., 2022).Machine learning algorithmically builds a statistical model based on a set of examples of some phenomenon (Burkov, 2019).In particular, supervised learning takes a set of examples with known outcomes to build a model that can predict the outcome of unseen examples, given a set of features (Burkov, 2019).Combining random forest and moving average models had 0.95 specificity and 0.54 sensitivity in predicting daily BRD status from feeding behavior and activity features from accelerometers (Bowen et al., 2021).The K-nearest neighbors model, that included AMF and accelerometer data, predicted BRD 6 d before diagnosis more accurately (0.96 vs. 0.52) than a model including only clinical signs scores and body temperature (Cantor et al., 2022).A convolutional neural network predicted sickness (BRD or ED indistinctively) based on AMF data with 0.83 sensitivity and 0.79 specificity for calves under restricted (6 L/d) or high milk (25 L/d) allowances.However, the feature importance differed between models specific to each milk allowance (Ghaf-fari et al., 2022).Under a restricted milk allowance, the number of daily unrewarded visits was the most important feature, while under a high milk allowance, drinking speed and daily milk consumption were the most important features (Ghaffari et al., 2022).
Even though machine learning models showed better specificity and sensitivity for disease prediction using AMF variables than self-starting cumulative sum charts, it is difficult to compare models across studies since they have used different data sets.There is also a potential benefit of combining multiple data sources.Model performance was improved by incorporating accelerometer data, since these features may capture changes in activity patterns independent of feeding behavior.Like the discrepancies in the associations of feeding behavior variables with disease, differences in feature importance to predict disease across studies can be partly attributed to differences in feeding plans.Overall, there is potential for AMF data to develop predictive models that assist disease detection in preweaned group-housed calves.

Social behavior in AMF systems.
Pre-weaned dairy calves are motivated to interact with other calves (Ede et al., 2022), and when housed in groups with AMF systems they benefit from social learning (de Paula Vieira et al., 2012) and social support (Bučková et al., 2022;de Paula Vieira et al., 2010).Interactions between calves, such as aggression and avoidance are considered agonistic while grooming, spatial proximity, and reduced aggression are affiliative (reviewed by Bouissou et al., 2001).Feeding behavior of calves in AMF systems can also be influenced by agonistic and preferential interactions with other calves in the group.
Cattle are gregarious animals and establish hierarchies and preferential bonds (reviewed by Bouissou et al., 2001).Calves establish preferential bonds based on their temperament and familiarity with other calves (Faerevik et al., 2007;Lecorps et al., 2019).When calves are introduced to a group with unfamiliar animals, they can experience stress and aggression (Bøe and Faerevik, 2003).In an experiment, calves that were introduced to an established group reduced their milk intake relative to the day before joining and received more competitive displacements than resident calves (O'Driscoll et al., 2006).Although on the day of mixing calves consumed less milk and performed fewer visits to the feeder, they returned to normal on the next day (O'Driscoll et al., 2006).However, group size and previous social experience could also influence the effect of introduction on feeding behavior.
Montes and Boerman: Literature review: Social and feeding behavior of pre-weaned calves Competition and aggression at the feeding station can cause calves to modify their feeding behavior.Calves experience aggressive interactions when they are introduced to a stablished group or in groups with elevated calf to feeder ratios (i.e., > 12 caves: 1 feeding station).Thus, grouping events and group density may also be important to consider when observing behavior patterns across different settings or over time.Understanding social preferences and competition experienced by pre-weaned calves can lead to better decisions on grouping and regrouping strategies in AMF systems.

Social networks in group-housed calves.
The interactions between group-housed AMF-fed calves can be represented by a social network and SNA can be used to understand the social structure.Animal social networks are relational data sets representing how each animal is connected to other animals (reviewed by Croft et al., 2008).Research indicates that SNA can quantify indirect relationships or associations between individuals, which allows for detecting patterns and structures within a population (reviewed by Beisner and McCowan, 2015).The properties of the network structure and individual's positions can be useful for the study of social phenomena in calves such as disease spread (Chen et al., 2013;Frelson et al., 2019) and social stability (Vázquez-Diosdado et al., 2023).However, continuous observation of animal behavior with traditional methods can be labor-intensive and subject to observer bias (Foris et al., 2019).In contrast, precision technologies keep continuous records of individual animals that with proper validation can be used to derive spatiotemporal associations (Rice et al., 2020).
Social network analysis is a promising way of looking at data collected from precision technologies.Recently, SNA has been used to study disease spread in calves with observed interactions (Frelson et al., 2019;Burke et al., 2022) and with interactions derived from automated data (Chen et al., 2013).In one study, contact interactions derived from position loggers were used to simulate disease spread in dynamic networks of weaned calves (Chen et al., 2013).The analysis of networks representing the contacts which facilitate Leptospira spp.transmission in one group of cows and 2 groups of weaned calves showed that individual attributes influenced the contact structure, and thus should be considered for prevention and control measures (Frelson et al., 2019).For example, in the group of older calves, there was a higher probability of contact between 2 calves if they were of similar age (Frelson et al., 2019).
In addition to understanding the role of social interactions in disease spread, SNA has also been used to identify associations between health and social behavior of calves.Vázquez-Diosdado et al. (2023) investigated the stability of social proximity networks built with data from position loggers in 5 groups of AMF-fed calves over a period of 76 d.Even though networks were not stable over time, the preference to interact with calves of similar age was maintained (Vázquez-Diosdado et al., 2023).The analysis also revealed that sick calves spent less time engaging in social interactions and had a lower position within the network (Vázquez-Diosdado et al., 2023).Similarly, calves infected during a disease challenge (Mannheimia haemolytica) were less connected in a network built from social lying interactions on the day of the challenge than the previous day (Burke et al., 2022).
Continuous monitoring of social interactions between group-housed calves has been possible with automated data collection with precision technologies.The analysis of dynamic networks can provide a deeper understanding of the relationships between the social structure of calves and events like disease.Simulated contact networks showed that calf attributes influence contact with other calves and disease transmission.Analysis on proximity networks showed that network position and changes in network structure can be related to health status.Future research with SNA on AMF-fed calves can lead to improvements in disease prevention and grouping strategies within AMF systems.While SNA show promise in improving our understanding of dairy calf group dynamics and disease transmission, more research is needed in this area to explore the role of social networks on calf performance within an AMF system.

CONCLUSION
Calves in small stable groups have less risk of disease and experience less agonistic interactions than calves in larger groups.Existing research has shown that AMF-recorded variables are associated with BRD and ED status and can be integrated with activity to build predictive models using machine learning algorithms.Changes in feeding behavior due to sickness are influenced by daily milk allowance and calves may only exhibit reduced consumption when they are fed larger milk volumes.Social network analysis on the structure groups of calves and its dynamics can lead to a better understanding of disease spread which could be incorporated into preventive practices.

Table 1 .
Montes and Boerman: Literature review: Social and feeding behavior of pre-weaned calves Summary of published research on disease and feeding behavior of calves in automated milk feeder systems

Table 1 (
Renaud et al., 2020.uirk andhed research on disease and feeding behavior of calves in automated milk feeder systems Details on the feeing plan have been omitted.3.Wisconsin scoring system the study referred to eitherMcGuirk, 2008 or McGuirk and Peek, 2014.4.Fecal scoring adapted fromRenaud et al., 2020.