Our vision is of a world where efforts to prevent and alleviate animal suffering are as efficient and cost-effective as possible. Where decisions across various domains (economy, environment, investment, purchasing) are also informed by their impact on the welfare of animals. Achieving this vision depends critically on properly measuring, and mapping, the magnitude of animal suffering across conditions, systems and species.
Ultimately, we would like to help transform the understanding of animal and human suffering, shifting it from an abstract concept to a scientifically measurable and extensively mapped phenomenon across all sentient beings. In doing so, we seek to help empower society (legislators, environmentalists, economists, advocates, funders, investors, consumers) to make decisions and take actions more closely aligned with ethical values that aim to minimize harm.
To inform efforts for the prevention and alleviation of animal suffering. To this end, our research is aimed at providing consumers, funders, researchers, investors, advocates and other decision-makers with a comparative, meaningful and easily interpretable metric (1) of the magnitude of suffering embedded in animal-sourced foods produced from different species and under different conditions and (2) of the impact of interventions, standards and welfare reforms. As a side-product of our research efforts, we also aim to provide robust and validated data for campaigns, negotiations and litigation cases, and to promote more targeted research and funding in the animal welfare sciences.
For measurement purposes, the term pain is used as a shorthand for any ‘felt negative (unpleasant) affective state’. More simply, anything that ‘feels bad’. States with a somatic origin are referred to as ‘physical pain’ (e.g. aches, hunger, injuries, thermal stress), while those related to the primary emotional systems are referred to as ‘psychological pain’ (e.g. fear, frustration, boredom).
The assessment of the time spent in pain (physical or psychological) of different intensities as a result of one or more harms of interest (e.g., diseases, deprivations). The use of a universal metric with biological meaning (time in pain) enables comparing and combining the cumulative load of negative experiences emerging from different environments, industry practices, production systems or interventions, and provides the basis for the development of welfare footprints of animal-sourced products.
Research and decisions aimed at improving the quality of life of animals are already guided by assumptions about their affective states and how they are modulated. For example, requirements for standards such as minimum space allowance, enrichment, outdoor access and stunning prior to slaughter (inputs) necessarily assume that these practices lead to improved well-being (i.e., reduce the intensity, frequency and/or duration of negative affective experiences). Many assessments also rely on indirect indicators of affective states, such as behavior and neurophysiology (outputs). Ultimately, any form of welfare assessment or attempt to improve animal welfare is inherently based on an assumed knowledge of what positively or negatively affects their inner experiences. The current framework is focused on inferring, using existing scientific evidence, parameters (time and intensity) natural to affective experiences. By atomizing the analysis into the effect of each welfare challenge on individuals, it ensures that the evidence and assumptions supporting the proposed parameter values are made explicit, helping drive targeted progress. Refocusing the welfare debate towards the study of affective experiences at the individual level (where experience occurs) also makes the assessment of animal welfare more tractable.
A flowchart with the main analytical steps is provided in our methods page and here. Briefly, the method relies on the assessment of the duration and intensity of the pain caused by each of the welfare challenges to which a population of interest is exposed (in the period of interest), along with the assessment of the prevalence of the challenge in the population. This is done in the following way:
Subjectivity in the categorization of pain intensity is a problem that permeates all intensity scales, both numerical and nominal. The advantage of nominal scales lies in removing one layer of complexity in the attribution of intensity to the pain experience: ‘mild’, with all its problems, is self-explanatory to most individuals, whereas ‘5’ must be explained and externally referenced. Use of discrete categories also avoids the problem of forcing numerical ratings that may not necessarily be linear into a linear scale, and prevents the ceiling effects that are frequently observed with numerical ratings of pain (where ratings close to the limits of the scale are constrained, increasing in intensity by smaller amounts than the intensity effectively experienced).
We sought to reduce the ambiguity of widely used terminologies (mild, moderate, severe) by using instead terms which, in addition to being semantically more specific, are qualified by specific references and criteria. Use of these criteria enables establishing more specific thresholds in the attribution of pain intensity that reduce the likelihood of misclassification, facilitating comparisons across conditions and individuals. For example, the pain of a fracture at the time of rupture matches the definition of a Disabling pain, not because this is the verbal description preferred by patients, but because it matches the definition of this intensity category: it captures nearly all the individual’s attention, sufferers are unable to perform other activities and strong analgesia is commonly required. Additionally, in this process the intensity and duration of the pain experience are completely disentangled (in some other approaches, the pain associated with a condition may be considered ‘severe’ because it lasts longer), further reducing variability in the classification process.
The number of categories was devised to represent an optimal balance between precision and clarity, with categories that are distinct enough to minimize the risk of misclassification.
Although pain inherently concerns individuals, we operationally accept that the collective welfare of the members of a population can also be determined. Measuring cumulative pain at the population level is also necessary to account for the heterogeneity in the exposure of population members to different challenges. For example, while lameness is experienced by a large fraction of broiler chickens, fatal cases of ascites are only experienced by a few. Therefore, measurement efforts must consider the prevalence of each welfare challenge, so that pain is determined for the average member of the population (which may not necessarily correspond to any real organism). At the population level, the time spent at each level of pain intensity by the ‘average population member’ as a result of each challenge is determined by multiplying it by its prevalence. For example, if a condition causes 10 hours of Disabling pain and 70% of the population are affected, then the average member of this population could be said to experience 7 hours of Disabling pain due to the condition. Measurements at the population level enable comparing the impact of different practices and conditions across demographics, geographies, and time
The method is receptive to all types of evidence. Any form of data or knowledge that has the potential to decrease the uncertainty associated with the potential intensity of a painful experience is considered as potentially valuable. Estimates of pain intensity are thus typically based on multiple lines of evidence, including neurophysiological indicators of pain, preference tests, pharmacological tests (e.g., effect of pain-killers), knowledge of existing pain-generating mechanisms in each case, observations of behavior and activity levels, and evolutionary reasoning. Pain is an attention-dominating state, a warning signal of threat (real or perceived) that must change behavior to protect the organism; the greater the threat, the more intense this signal should be. Accordingly, more unpleasant sensations should in general be more disruptive and engage more attention. Four discrete categories of intensity are defined based on these criteria (Annoying, Hurtful, Disabling and Excruciating). For each welfare challenge analyzed, evidence of the pain it causes is compared against each category’s definition to inform the probability that the pain is of each intensity.
The greater the strength and quality of the evidence that a given painful state matches the criteria defining one of the intensity categories, the greater the probability attributed to that category. Since the main adaptive value of pain is to prompt protective behavioral responses, behavioral indicators of pain are particularly valuable. Yet, the specific probability assigned to each category is still subjective. In recent developments of the framework, this subjectivity is constrained by the use of predefined and relatable probability categories. Specifically, for ‘each’ of the pain intensity levels, we ask: “Based on existing evidence, what is the likelihood that pain, for this welfare problem at this stage, is of the following intensity?”. Answers are limited to the following possibilities: CERTAIN (there is a 100% probability that pain is of this intensity), VERY LIKELY (there is a 90-99% probability that pain is of this intensity), LIKELY (there is a 60-89% probability that pain is of this intensity), POSSIBLE (there is a 40-59% probability that pain is of this intensity), UNLIKELY (there is a 10-39% probability that pain is of this intensity), VERY UNLIKELY (there is a 1-9% probability that pain is of this intensity), IMPOSSIBLE (there is 0% probability that pain is of this intensity). Naturally, (1) each choice must be as thoroughly justified as possible and (2) there are constraints on the answers. For example, it is not possible to estimate, for a given welfare problem at a point in time, that pain is both “certainly” Excruciating and “certainly” Disabling (the sum of probabilities across all four pain intensity levels cannot exceed 100%).
In the unlikely event that no information was available, maximum uncertainty would be assumed, with the same probability (20%) attributed to each of the intensity categories (no pain, Annoying, Hurtful, Disabling, Excruciating). So far, however, we have not come across any scenarios in which information is completely unavailable.
Validation (or falsification, in Karl Popper’s sense) of estimates of the intensity and duration of affective states using direct empirical evidence is not yet possible (constraints imposed on the validation of hypotheses empirically are also present in other scientific areas, such as astronomy). Because the assessment of affective states still relies on indirect evidence, the attribution of probabilities of pain intensity and duration to each temporal segment of a painful event is an (always open) exercise of presenting evidence, from as many relevant lines of enquiry and sources as possible, to justify the values proposed. Typically, hypotheses on parameter values are reviewed by other specialists in the field. But estimated parameter values are always treated as hypotheses, each associated with an uncertainty level (represented by subjective confidence intervals), which may be reduced (1) as more evidence becomes available and/or (2) greater consensus is reached towards proposed values by independent evaluators. In this way, proposed figures are continuously updated and strengthened.
Estimates of intensity and duration associated with each welfare challenge are inputted in a notation tool, the Pain-Track, from which cumulative time in pain of each intensity are calculated. These estimates are based on existing scientific evidence (e.g., behavioral, neurophysiological, immunological indicators of welfare, knowledge of neurology, evolutionary reasoning). Thus, calculations of Cumulative Pain take full advantage of relevant existing evidence, while being limited only by its accuracy. In other words, estimates of Cumulative Pain are as accurate as the integration of existing practical and theoretical knowledge of animal welfare allows.
A single, aggregate metric of welfare loss would be more convenient for practical purposes. However, evidence to support inferences on the mathematical equivalence between the intensity categories is not yet available, so any integration exercise would still require relying only on subjective assessments of such an equivalence. Additionally, the validity of balancing different levels of suffering in comparisons involving different individuals or populations is not clear. For example, estimates of Cumulative Pain in a population with individuals enduring intense pain could be similar to that of a population where no individual suffers intense pain, but a larger fraction of individuals experience milder pain for a sufficiently long time. We believe that, so far, analyzing estimates of the total time spent in pain of each intensity represents a more accurate and transparent approach, grounded in explicit parameters with biological meaning. It also enables the pursuit of different priorities, such as relieving the worst kinds of preventable suffering (those associated with the Disabling and Excruciating categories of pain). Moreover, presentation of the disaggregated estimates enables their integration by audiences with different views on the weight that should be given to each level of pain intensity (e.g. here and here).
A time-based equivalence between levels of pain intensity, based on scientific evidence, is not yet available. So far, the answer to this question is more of a personal, moral or philosophical nature. Because the focus of the Welfare Footprint Project is the production of scientific and evidence-based knowledge, we have opted for keeping the results of the analytical process untransformed, and leave the subjective weighing of pain intensity categories for individuals and institutions that will make use of them, and which may hold different views on what those weights should be. Still, it is possible to determine how much worse a given pain level would need to be for two situations to be considered equivalent in terms of suffering. For example, say that a typical bird kept in system ‘A’ experiences about 1,000 hours of Hurtful pain, whereas another living in a system ‘B’ experiences about 5 hours of Disabling pain. For the systems to be considered equivalent in terms of suffering, the Disabling intensity would need to be 200 times worse than the Hurtful intensity during all the time it is felt (1,000/5). If one finds this figure (200) too high, then system ‘A’ is judged to be worse than ‘B’. Otherwise, system ‘B’ is judged as better.
So far, estimates of Cumulative Pain have been based on the operational assumption that multiple concurrent challenges have an additive effect on experience: namely that their effects when experienced simultaneously are best represented by the sum of their isolated effects. This solution works well for the purpose of distinguishing the impact of welfare challenges across scenarios or production systems.
However, further refinements are being discussed to estimate (1) the time individuals spend in negative affective states when experiencing concurrent challenges of similar intensity, and (2) the time individuals spend in negative affective states when experiencing concurrent challenges of different intensity. In the latter case, when the pain from a challenge is so intense as to grab all or nearly all the individual’s attention (as is the case of Disabling or Excruciating pain), pain of lower intensity from other concurrent harms will be likely overridden, and should not be considered.
A much more thorough discussion is available here. We would greatly appreciate any feedback that can help inform our decision-making in this regard.
This phenomenon happens because individuals may experience more than one source of pain at the same time.
The Cumulative Pain framework estimates the time individuals typically are exposed to pain of each intensity: one hour of Hurtful pain means that an individual was exposed to a source of Hurtful pain for one hour. In its current form, the framework thus represents two simultaneous sources of Hurtful pain, each lasting an hour, as two hours of Hurtful pain. This choice offers the benefit of simplicity, and eliminates the need for more complex assumptions about the type of additivity in the subjective perception of pain (for which knowledge is not yet available) when individuals experience welfare challenges that overlap in time .
However, estimates computed in this manner often raise the question of how an individual can experience more hours of pain than there are hours in a day, or even a lifetime. For instance, consider a fictitious case: an animal spends her life with two ailments: (1) a mild “itch” (of an Annoying intensity) and (2) a mild thermal discomfort (of an Annoying nature too). She lives for 60 weeks, and pain is only felt when awake (16 hours/day, totaling 6,720 hours awake). In this case, she experiences 6,720 hours of Annoying pain due to the itch and 6,720 hours of Annoying pain due to thermal discomfort. The total of 13,440 hours of Annoying pain is longer than her life. This is a by-product of a technical choice in the approach.
One potential solution to this issue involves incorporating the degree of attention allocated to each pain intensity category into Cumulative Pain calculations. In other words, instead of estimating the duration an individual is subjected to pain (i.e., the entire period during which an individual is exposed to an unpleasant experience), one would estimate the time spent attending to pain, namely the duration of those moments an individual actively focuses on or is aware of this sensation. A full discussion of the share of attention that may be dedicated to pain, in each intensity category, is available here. Time estimates calculated in this manner will be typically shorter, as individuals may not be continuously aware of their discomfort or may actively choose to focus on other aspects of their experience part of the time.
For instance, if we assume that an individual is aware of Annoying pain only 25% of the time he is exposed to a discomfort of this intensity (in line with the definition of Annoying pain), then each hour of Annoying pain exposure should be converted to 15 minutes of actual time spent in Annoying pain. Consequently, two simultaneous sources of annoying pain lasting one hour would be translated into a maximum of 30 minutes of annoying pain.
By incorporating attention allocation into the framework, we can more accurately estimate the actual time spent aware of a negative affective state. This also resolves the apparent contradiction of cumulative pain estimates exceeding a lifetime when individuals are exposed to more than one source of pain simultaneously. However, the decision to incorporate this refinement depends on the reliability of estimates for the share of attention allocated to painful experiences of varying intensities – for which evidence is not yet available. Adopting provisional values that may change in the future could harm the comparability of analytical results. We are therefore gathering more feedback on this issue before incorporating this refinement. Any input or contribution are thus greatly appreciated.
Considering the importance, and greater urgency, of preventing animal suffering, the method has so far given priority to the quantification of negative affective states (an approach in line with suffering-focused ethical views). A focus on the reduction of suffering also stems from observations that negative experiences have a larger impact on well-being than positive ones (e.g., as improvements in positive states are frequently bounded, quickly reaching diminishing returns, as well as lost to hedonic adaptation, whereas adaptation to pain is unlikely). In the case of humans, this perspective forms the basis of policies guiding efforts at poverty alleviation and disease mitigation.
It is also important to highlight that it is reasonable to assume that positive welfare is precluded, either partially or completely, in animals experiencing negative affective states, particularly the more intense ones (this assumption is in fact part of the defining criteria of the four intensity categories considered in the Cumulative Pain framework). In other words, understanding first the amount of time individuals spend focusing on pain is necessary if we are to understand the time remaining for the experience of positive states. A more thorough discussion of this topic is available here.
Positive experiences can then be incorporated into analysis by expanding the framework to include estimates of the time spent in positive affective states. In the latter case, the same analytical steps would apply. For example, a “Pleasure-Track” could be used to describe hypotheses on the temporal evolution and intensity (e.g, in categories from Joy to Bliss) of positive states, which would be then used to estimate the Cumulative Time in Pleasure at the individual and population levels. Development of this refinement by other researchers and analysts would be most valuable.
At any time, the subjective well-being of individuals depends on an integration between positive and negative affective states, namely an yet undetermined subjective calculation of pleasure relative to suffering that may be positive or negative. Integration of positive and negative states over a longer time frame is also at the core of the notion of a “life worth living” or “good life”, or one where the surplus of pleasure over suffering is positive. Although such an integration exercise would be more truthful to the range of experiences an animal endures, it is still hindered by the challenges to establish a mathematical equivalence between positively and negatively valenced states. For example, the extent to which positive and negative experiences are morally symmetrical, or pleasure can compensate for time in pain, is far from consensual (see previous question). Even among humans, who can verbalize their preferences, whether time in pain of a certain intensity can be compensated for time of healthy and joyful life will often come down as a personal and moral choice. What a person (or group) may consider an acceptable trade-off, others might condemn. Additionally, compensation between positive and negative experiences is particularly difficult in cases of extreme suffering: would there be any magnitude of pleasure that would compensate for torture-like experiences? Another drawback of the integration of positive and negative effects are the implications for the analysis at the population level. If welfare is measured as a surplus of pleasure over suffering, would severe suffering endured by a few individuals be compensated by barely positive welfare endured by a sufficiently large number of individual
The framework can be applied to any sentient species, including humans. Because the metric used is based on a phenomenon relevant to all sentient beings (time in pain of different intensities), it can be applied to virtually any sentient species to assess any context of relevance.
The Cumulative Pain metric can be used to estimate and compare the magnitude of suffering associated with different welfare challenges experienced by wild animals under different conditions and even by different species. In addition to enabling quantitative comparisons, an inventory of the magnitude of suffering associated with different challenges that wild animals typically endure can also promote awareness about wild animal suffering, help identify hotspots of suffering in the wild and, importantly, contribute to the design strategies to reduce suffering in the wild without burden shifting. An example of the application of the framework to wild animals is available here.
Do three hours of Disabling pain in mammals represent the same suffering as three hours of Disabling pain in fish, or shrimp? The answer to this question is not simple. The extent to which different species differ in their capacity for affective experiences is one of the biggest questions of present and past times (the results of a major research effort to answer this question, the Moral Weight Project, is available here). Yet, it is one that must be addressed if Welfare Footprints are to be compared across species. The Welfare Footprint Project team will be soon joining efforts in this line of enquiry to investigate what might be at least workable solutions to this question.
That said, this challenge does not preclude use of the same assessment methodology across species. Cumulative Pain estimates are based on universal parameters that should apply to any sentient organism: duration and intensity of affective experiences. Likewise, the criteria defining the categories of intensity are guided by measurable consequences of pain and general evolutionary principles that should also apply to most sentient creatures, such as how disruptive the experience is, how much attention it demands, whether it allows for positive states, how individuals trade-off pain for other experiences (positive or negative) and the extent to which it can be mitigated by pharmacological interventions (e.g., painkillers). For example, although the measurable manifestations of pain vary across species, the experience of Disabling pain for a pig should – in principle – be as attention-demanding and distressing as for a fish. Similarly, species may experience pain of similar intensity even if they differ in the ‘type’ of affective experience they endure as a result of idiosyncrasies in their anatomy, neurophysiology and cognition (e.g., grief is unlikely to be a universal emotion, but pain of intensities similar to those triggered by grief is likely experienced by multiple species due to other welfare challenges). Because the focus of the Cumulative Pain framework is on the intensity of pain, not type of pain, comparisons become easier.
This transformation can be applied to Cumulative Pain estimates by anyone who wishes to do so. Still, we focus on absolute time in pain since: (1) conversion of absolute time in pain to a proportion of lifetime in pain is predominantly driven by a desire to compare lifetime spent in unpleasant states with lifetime spent in neutral or positive states. Yet, the extent to which the latter can compensate for the former, or the validity of such compensation at the population level, is far from consensual (see question “how would positive experiences compensate for negative ones?”); (2) the extent to which 10 hours of Disabling pain in a lifetime of 100 hours (10%) is the same as 1,000 hours of Disabling pain in a lifetime of 10,000 hours (10%) is not clear to us; (3) in practical terms, such a conversion exercise would only apply to analyses where the effect of each and every adverse event over the entire lifetime of individuals is considered, to ensure that the time assumed to be in neutral/positive states is completely devoid of negative experiences (the type of analysis not yet conducted by the WFP team); and (4) when multiple harms are experienced simultaneously, estimates of time in pain can be longer than the lifespan of individuals.
No. The Cumulative Pain framework does not assign a negative value to death, only to negative affective experiences (anything that ‘feels bad’).
A meaningful and universal metric of the loss of welfare embedded in different animal-sourced products, practices and production contexts has different practical uses for different audiences:
