At the heart of sentencing: exploring whether more compassionate delivery of sentencing remarks increases public concern for people who offend

Abstract

Compassion has the capacity to change how we think and feel about people who offend, enabling us to understand individual and systemic causes of criminality and whether, and in what circumstances, desistance is possible. Across two experiments, our research examined whether a more compassionate sentencing delivery, firstly, in written sentencing remarks and, secondly, in videoed sentencing remarks, stimulated more concern for sentenced offenders amongst members of the Australian public. Our results suggest that it is possible to alter the features of a written or orally-delivered sentence, so that it is recognisably more compassionate. Further, engagement with compassion-enhanced sentencing remarks altered criminal justice spending preferences, reducing the proportion of the criminal justice budget that the public believed should be spent on imprisonment and increasing that to be spent on rehabilitation.

Introduction

Australian imprisonment rates are rising, with devastating impact on Indigenous peoples and other marginalised groups, such as those suffering from mental illness. This punitive trajectory is unsustainable, both morally and economically. However, to date, reasoned arguments for changing the course have yielded little support. According to Maruna and King, ‘[j]ustice is, at its heart, an emotional, symbolic process, not simply a matter of effectiveness and efficiency’ (2008, p. 347). Thus, understanding the trajectory of criminal justice law reform requires an engagement with emotions (Karstedt, 2002; Loader, 2011) – particularly fear – as key drivers of punitiveness, which cause us to turn away from those who are the subject of criminal sanction, othering them as monsters (Kennedy, 2000). This can, paradoxically, result in law reform that reduces public safety, rather than increasing it (Bartels & Hopkins, 2021).

By contrast, more reasoned criminal justice reform requires us to turn towards those facing sanction, as human beings, to understand the individual and systemic causes of their criminality and consider whether and in what circumstances desistance is possible. This turning towards describes a compassionate approach, in which we are ‘motivated to engage with suffering, to stay with it and understand its causes in a non-judgemental way [and] skilfully work towards the alleviation and prevention of this suffering and its causes’ (Gilbert & Choden, 2014, p. 174). Several studies have found empirical support for a relationship between empathy and punitiveness, with more empathic individuals less likely to support capital punishment (Unnever & Cullen, 2009) and more likely to hold favourable attitudes toward offender rehabilitation (Mackey et al., 2006).

The value of compassion as a foundational psychology for actors working within the criminal justice process is well recognised, including amongst judicial officers tasked with sentencing offenders (Hopkins et al., 2019; Roach Anleu & Mack, 2013). Compassion may also have a larger role to play in reorienting law reform, from its increasingly punitive trajectory, towards a more understanding and rational approach. This raises the question of whether, and how, compassion can be cultivated within the broader public, influencing the ‘law-and-order’ conversation. While working directly with members of the public to develop and cultivate compassion shows great promise (e.g. Goleman & Davidson, 2017), the reach of such interventions is limited and stands against the broad reach and narrative power of the press and social media. This makes us consider whether engaging members of the public with a more compassionate narrative about people who offend, via compassion-enhanced sentencing remarks, delivered compassionately by sentencing judges, has the capacity to stimulate a more compassionate and less punitive response. This is a strategy that goes directly to the heart: identifying, engaging with and promulgating compassion towards people within the criminal justice system (CJS), in pursuit of a more compassionate and reasoned policy conversation (Bartels & Hopkins, 2021).

A key site for the emergence of more compassionate relationships with those who offend is at the point of sentencing. Every day, across Australia, judges and magistrates are routinely engaged in the process of sentencing. The sentences themselves, together with the remarks that set out the reasons for sentence, are delivered orally, mostly in open court. They are delivered in the presence of the person being sentenced, any victim who chooses to attend, the family of those affected, interested members of the public and the press. In the higher courts, which sentence a smaller number of people, who have typically committed more serious offences, these sentences are also delivered in writing and published on publicly available databases. In these published sentencing remarks, a characterisation of the person reaches the public, albeit usually through the filter of media reporting. This narrative can be more or less compassionate. That is to say, sentencing remarks can vary in how they express compassion towards the person facing sentence and those affected by the offending, including any victim(s). The oral delivery of these remarks also has the potential to display more or fewer features of compassion.

In the two studies reported here, we sought to identify these features of compassion, measure whether observers of compassion-enhanced or standard sentencing remarks perceived the former to be more compassionate than the latter, and then test the hypothesis that engaging with a more compassionate delivery of a sentence stimulates a more compassionate and less punitive response to those who offend. We also aimed to investigate whether engagement with a more compassionate narrative would alter the public’s preferences with respect to criminal justice spending – for example, on incarceration.

These studies contribute to the growing literature on the role of emotion in law and therapeutic jurisprudence, which recognises law as a social force that can produce therapeutic or anti-therapeutic outcomes (Winick & Wexler, 2003) and that the law, legal processes and legal actors can play a role as therapeutic agents (Wexler & Winick, 1996). It draws inspiration from the Australian Feminist Judgments Projects (see Douglas et al., 2014) and many examples of what we regard as compassionate sentencing remarks (see, e.g. R v Armitage, 2015; DPP v Pell, 2019).

To conduct these experiments, it was first necessary to settle upon a robust definition of compassion, supported by the psychological and therapeutic jurisprudence literature, to facilitate the compassionate rewriting of standard sentencing remarks. An attendant focus was the need to identify features of more or less compassionate oral delivery of these remarks, in terms of posture, voice tone and pace, expression and the like. Second, to robustly measure the influence of including compassionate features with the potential for real-world application, it was necessary to ensure that the features included in the rewrite and oral delivery were consistent with the lawful exercise of the sentencing discretion. In other words, we needed to ensure that the changes made were changes that could be made by a sentencing judge or magistrate, without compromising the sentencing task.

A definition of compassion

There is broad consensus that compassion involves a sensitivity to suffering and motivation to act to alleviate it (e.g. Dalai Lama, 1995; Gilbert & Choden, 2014; Strauss et al., 2016). However, there is no clear consensus on its precise definition, or its essential attentional, affective, cognitive and behavioural elements. This presents challenges for the study of compassion, its measurement and the evaluation of whether an intervention is effective in increasing compassion (Strauss et al., 2016).

In our case, settling on a robust definition of compassion that identified its core elements was essential, to enable us to alter the sentencing remarks and their delivery, to increase the extent to which they exhibited features of compassion. To address this challenge, we settled on the Strauss et al. (2016) definition, which was based on a systematic review of the literature:

1. Recognising suffering is an explicit or implicit feature of all definitions of compassion (Strauss et al., 2016), which can be understood as noticing (Kanov et al., 2004), a sensitivity to, or an awareness of suffering (Gilbert, 2009).

2. Understanding the universality of human suffering contains understanding of ‘the universality of pain in human experience’ (Feldman & Kuyken, 2011, p. 145) and common humanity (Pommier, 2010), which engages a sense of ‘[t]here but for the grace of God, go I’ (Strauss et al., 2016, p. 17), non-judgement and universal application. This understanding is critical to ensuring that compassion is not ‘selectively doled out’, based on in-group bias (Bandes, 2017, p. 184), but instead applies to all people – here, for example, those who offend and victims alike – in their full humanity and dignity as equals (Hopkins et al., 2019).

3. Feeling empathy for the person suffering and connecting with the distress (emotional resonance) is an explicit element of all definitions of compassion and may be expressed as feeling, meeting, being open to, or moved or touched by the suffering in the person’s experience (Dalai Lama, 1995; Feldman & Kuyken, 2011; Gilbert, 2009; Kanov et al., 2004; Pommier, 2010; Strauss et al., 2016). Connecting with the distress entails a physical or affective response, in which the pain of the sufferer’s experience is felt, and shared, by the person witness to it. This can be understood as the vicarious experiencing of another’s emotions (Strauss et al., 2016).

4. Tolerating uncomfortable feelings aroused in response to the suffering person (for example, distress, anger or fear), so as to remain open to and accepting of the sufferer is also a well-recognised and critical element (Strauss et al., 2016). This entails a capacity to stay with the arising emotional distress and thereby stay turned towards the sufferer and their experience, rather than becoming overwhelmed and turning away (Hopkins & Bartels, 2019).

5. Motivation to act or acting to alleviate suffering is again an explicit element of all definitions of compassion (Strauss et al., 2016) and is frequently singled out as the key feature that distinguishes compassion from empathy, such that compassion incorporates, but is more than, empathy (Gilbert & Choden, 2014; Hopkins & Bartels, 2019; Strauss et al., 2016). This motivation to act may variously be described as wanting to help, a commitment to relieve, or a desire or wish to alleviate and includes acting or responding to alleviate and prevent suffering and its causes (Dalai Lama, 1995; Feldman & Kuyken, 2011; Gilbert, 2009; Kanov et al., 2004; Pommier, 2010; Strauss et al., 2016).

It is important to say something about what compassion is not. Compassion is not pity, feeling sorry for or looking down on a person. Pity entails a paternalism and superiority, which denies the universality of human experience and the inherent equality of human beings (Hopkins & Bartels, 2019). Moreover, pity does not require or motivate action to alleviate suffering and may in fact imply that the person suffering is unworthy of help (Strauss et al., 2016). Similarly, compassion is not mercy. Whilst mercy may entail action, it too is suggestive of paternalism that denies the universality of human experience (Hopkins & Bartels, 2019). Finally, and of particular importance in the context of sentencing, compassion is not soft. Working skilfully to alleviate and prevent suffering may entail, amongst other things, sentencing a person to imprisonment or imposing other restraints to protect the community, or to achieve other recognised sentencing purposes (Hopkins & Bartels, 2019).

Sentencing as a space for the expression of compassion

A sentencer is required to consider all of the relevant circumstances in tailoring a sentence that achieves a balance of competing purposes, ‘protection of society, deterrence of the offender and others who might be tempted to offend, retribution and reform’ (Veen v R (No 2) (1988), p. 475). These purposes now exist in statutory form in all Australian jurisdictions, with some extension and refinement, including specific reference to the need to recognise harm to the victim and community (see e.g. Crimes (Sentencing) Act 2005 (ACT), s.7).

The process whereby a sentencer arrives at a sentence is discretionary. There is no single correct sentence arrived at mathematically. Rather the sentencer must identify all relevant features of the offence and offender, discuss their significance and make a judgement as to the appropriate sentence, in light of these features (Wong v R (2001); Markarian v R (2005)). Sentencing remarks made by a judge or magistrate must reveal the reasons for the decision in sufficient detail to enable the decision to be understood by the sentenced person, any victims, appeal courts and the public (Markarian v R (2005)). Aside from a requirement that these reasons be comprehensible to a lay listener or reader (Taylor v R [2018]), their form and the manner of their delivery is a matter for the sentencer.

Accordingly, the words and delivery of a sentencing decision may exhibit more or fewer compassionate features, understood by reference to the elements discussed above. This enabled us to alter a ‘standard’ sentencing decision and its delivery, to increase the presence of these features. The specific enhancements for each experiment are discussed below. Critically, the changes we made could be made by any sentencer, without compromising the exercise of their discretion. Further, it is important to note that the focus was on enhancing recognisable features of compassion in the text and delivery. The experiment was not intended to suggest that a sentencer whose remarks do not include these features is lacking in compassion or even less compassionate than one whose remarks include these features.

The current study

Though the role of compassion in law is receiving increasing attention, there has been no empirical testing of whether the expression of compassion towards those being sentenced (in addition to victims) has the capacity to influence public punitiveness. The aim of these experiments was, therefore, to examine whether compassion-enhanced sentencing remarks alter the public’s compassion toward those being sentenced or criminal justice attitudes (including punitiveness and preferences for criminal justice spending). We also aimed to test the effect of compassion on public perceptions of offenders as in- or out-group members and their capacity to experience various emotions. Finally, we examined the impact of compassion on participants’ views on the adequacy of the sentence imposed.

This project obtained ethics approval from the Australian National University. Using an experimental design, two experiments were conducted to test the hypotheses that sentencing remarks with more compassionate elements will increase the public’s compassion toward offenders, reduce their punitiveness and lead them to allocate criminal justice spending more to offender and rehabilitative-focused initiatives as opposed to more punitive spending options. In Experiment 1, an online Australian community sample was presented with written sentencing remarks, with approximately half receiving ‘standard’ sentencing remarks and the other half receiving altered remarks containing increased elements of compassion toward the person being sentenced (and others). In Experiment 2, a different online Australian community sample were presented with a video of sentencing remarks being delivered, with one version again altered to contain a more compassionate approach.

Measures

Across two experiments, we tested the effect of sentencing remark condition (standard versus compassion-enhanced remarks) on seven dependent variables. The same measures for all dependent variables were used for both experiments. Several covariates were also included in the analysis. Additional detail on the dependent variables and covariates are provided below, while the specific manipulations made to the sentencing remarks are discussed in the description of the method used for each experiment.

Scale for compassion and measuring importance of compassion for offenders

In both experiments, we adopted Strauss et al.’s (2016) conceptualisation of compassion as discussed above. To incorporate the five elements and reflect a criminal justice context, we slightly adapted the wording used by Strauss et al. to describe these core dimensions to create five statements that measure compassion for people who offend (for the survey instruments, see Supplementary Appendices A and B).

For the questions to detect compassion (i.e. the manipulation check), participants were asked whether they detected the five dimensions of compassion (see Table 1 for Experiment 1 and Table 7 for Experiment 2), with the question stem: ‘In the sentencing remarks that you read/watched, please indicate if the judge demonstrated the following approaches in their sentencing remarks’. These questions were intended to determine whether participants had identified the more prominent features of compassion included in the compassion-enhanced condition. Here, we expected to find a significant difference between the two groups, with participants in the compassion-enhanced condition expected to show greater recognition of the compassionate elements of the remarks.

Table 1.

Experiment 1: T-test results on participants’ mean scores on manipulation check questions.

Did the judge…	Group		t	Effect size(η2)
	Standard (n = 104)M (SD)	Compassion-enhanced (n = 102)M (SD)
…recognise the suffering of the offender?	3.59 (0.89)	3.92 (0.95)	2.61*	.032
…understand that the offender (as well as victims) has suffered?	3.72 (0.81)	3.92 (0.96)	1.62	.013
…show emotional responses toward the offender’s suffering?	3.38 (0.90)	3.60 (1.03)	1.66	.013
…acknowledge uncomfortable feelings about the offender’s suffering?	3.10 (1.01)	3.25 (1.08)	1.02	.005
Was the judge motivated to alleviate the suffering of the offender?	3.46 (0.87)	3.64 (0.95)	1.38	.009

*p < .05.

Table 7.

Experiment 2: T-test results on participants’ mean scores on manipulation check questions.

Did the judge…	Group		t	Effect size(η2)
	Standard (n = 119)M (SD)	Compassion-enhanced (n = 103)M (SD)
…recognise the suffering of the offender?	3.86 (0.98)	4.02 (0.92)	1.27	.007
…understand that the offender (as well as victims) has suffered?	3.93 (0.93)	3.92 (0.89)	–0.08	.000
…show emotional responses toward the offender’s suffering?	3.36 (1.22)	3.46 (1.06)	0.61	.002
…acknowledge uncomfortable feelings about the offender’s suffering?	3.38 (1.21)	3.50 (1.07)	0.76	.003
Was the judge motivated to alleviate the suffering of the offender?	3.46 (1.10)	3.75 (0.92)	2.09*	.019

*p < .05.

The second component measured participants’ views about whether compassion ought to be shown to people who offend. In this part of the survey, participants were presented with a series of statements, both before and after reading/watching the sentencing remarks, and were asked to rate their level of agreement on a 5-point Likert-type scale (1 = strongly disagree to 5 = strongly agree). By presenting these items twice, we were able to measure what effect, if any, the version of sentencing remarks (written in Experiment 1 or audio-visual in Experiment 2) had on participants’ compassion toward the person being sentenced.

For the pre-remark measure of the importance of extending compassion to people being sentenced, we asked a question with this stem: ‘You will now be asked some questions about your feelings toward people who are sentenced after being convicted of offences. Please rate on the scale provided if you agree or disagree with these statements’ (see Tables 2 and 8 for Experiments 1 and 2, respectively). For the post-remark measure, we asked this question again, this time asking participants to reflect specifically on the person they had read about (Experiment 1) or watched (Experiment 2). Here, we used the following stem: ‘When thinking about the sentencing of the offender in the video/you read about, please rate on the scale provided if you agree or disagree with these statements’.

Table 2.

Experiment 1: T-test results on participants' mean difference scores on compassion toward offenders.

	Group								t	Effect size (η2)
	Standard				Compassion- enhanced
	N	Pre- scoreM (SD)	Post-score M (SD)	Post-pre difference M (SD)	N	Pre- score M (SD)	Post- score M (SD)	Post-pre difference M (SD)
We should recognise the suffering of offenders	94	3.12 (1.11)	3.16 (1.02)	0.04 (0.81)	95	3.05 (1.12)	3.16 (0.98)	0.10 (0.98)	0.48	.001
We should understand that both offenders (as well as victims) suffer	94	3.20 (1.00)	3.39 (0.10)	0.19 (0.86)	95	3.32 (1.10)	3.48 (0.99)	0.17 (0.74)	–0.20	.000
We should feel emotional responses toward offenders who suffer	94	2.68 (1.03)	2.78 (1.09)	0.09 (0.90)	95	2.74 (0.96)	2.84 (1.03)	0.10 (0.78)	0.08	.000
We should acknowledge uncomfortable feelings we have about offenders who suffer	94	3.14 (0.98)	3.01 (1.04)	–0.13 (0.69)	95	3.24 (1.05)	3.14 (1.03)	–0.10 (0.89)	0.19	.000
We should be motivated to act to alleviate the suffering of offenders.	94	2.63 (1.03)	2.86 (1.13)	0.23 (0.66)	95	2.79 (1.03)	2.99 (1.03)	0.20 (0.71)	–0.34	.001

 

Table 8.

Experiment 2: T-test results on participants' mean difference scores on compassion toward offenders.

We should…	Group						t	Effect size (η2)
	Standard (n = 103)			Compassion-enhanced (n = 97)
	Pre- score M (SD)	Post- score M (SD)	Post–pre difference M (SD)	Pre- score M (SD)	Post- score M (SD)	Post–pre difference M (SD)
… recognise the suffering of offenders.	3.32 (0.97)	3.45 (0.98)	0.1262 (0.85)	3.05 (1.04)	3.53 (0.85)	0.4742 (0.87)	2.87**	.040
… understand that both offenders (as well as victims) suffer.	3.40 (0.94)	3.65 (1.03)	0.2524 (0.79)	3.20 (1.02)	3.60 (0.93)	0.4021 (0.90)	1.25	.008
… feel emotional responses toward offenders who suffer.	3.14 (0.98)	3.21 (1.035)	0.0777 (0.67)	2.89 (1.019)	3.21 (0.978)	0.3196 (0.98)	2.02*	.021
… acknowledge uncomfortable feelings we have about offenders who suffer.	3.42 (0.96)	3.44 (1.05)	0.0194 (0.74)	3.18 (0.94)	3.42 (0.94)	0.2474 (0.92)	1.917	.018
… be motivated to act to alleviate the suffering of offendersa	3.00 (0.99)	3.24 (0.99)	0.2427 (0.68)	2.81 (0.95)	3.27 (0.92)	0.4470 (0.84)	1.88	.018

Note: p < .05; ** p < .01; ***p < .001.

^an = 96.

We hypothesised that larger movements toward recognition of the importance of showing compassion for people who offend would be found amongst participants in the compassion-enhanced condition than amongst those in the standard condition.

Criminal justice budget allocation

Taking inspiration from Cohen et al.’s (2006) study of the American public’s criminal justice spending priorities, we asked participants to imagine themselves as their state/territory’s Premier/Chief Minister. They were told they had $100 per household to spend on combatting crime. After reading the judge’s sentencing remarks, participants were asked to decide how they would spend this $100, selecting from five options:

1. More prisons;

2. More police on the street;

3. More drug and alcohol treatment programmes for offenders;

4. More mental health treatment programmes for offenders; or

5. More programmes aimed at the welfare of victims.

Although participants were not told this, we consider the first two options to be punitive, Options 3 and 4 to be rehabilitative and Option 5 to be neutral.

Punitiveness toward offenders

To measure participants’ support for harsher treatment of offenders, we presented them with seven statements on the adequacy of punishments imposed. These items have been used in previous Australian research to measure public punitiveness (e.g. Dodd, 2018) and included: ‘people who break the law should be given stiffer sentences’ and ‘rehabilitation is not taken seriously by criminals’. A 5-point response set (1 = strongly disagree to 5 = strongly agree) was used for these items.

Like the compassion measures, these items were presented to participants once before reading/watching the sentencing remarks and again following the remarks. We anticipated that, comparing participants’ mean difference in pre-punitiveness and post-punitiveness scores, we would find significantly greater reductions in desires for harsher punishment amongst participants in the compassion-enhanced condition than amongst those in the standard condition.

Group membership

We asked participants two questions regarding whether they viewed the offender as an ‘in-group’ member, being a social group of which a person psychologically identifies as being a member (or person somewhat similar to themselves), or ‘out-group’ member – namely, a social group with which the individual does not identify. Participants’ level of agreement was again measured using a 5-point response set. We expected that participants who read/viewed the compassion-enhanced sentencing remarks would be more likely to view people who offend as an in-group member.

Attribution of emotions

Related to the previous question, participants were also asked to rate the likelihood that the offender could show different emotions. Following work on ‘infrahumanisation’ effects by Leyens et al. (2000), where the attribution of primary and secondary emotions to a target person is used as a proxy for whether that person is considered to be an in-group or out-group member, arguably deserving of compassion, as a fellow in-group member or even human, we classified the rated emotions as primary (consisting of anger, disgust, fear, joy, sadness and surprise) or secondary (affection, admiration, conceit, nostalgia, pride and remorse) emotions. According to Leyens et al. (2000), people should more easily associate their in-group member with secondary emotions than an out-group member, assuming that attribution of primary emotions only, with few attributions of secondary emotions, suggests the person is thought to be an out-group member, or, even, more like an animal than a fellow in-group human. Moreover, infrahumanisation researchers even find that, in some attribution tasks, participants deny the possibility that the target person can express secondary emotions at all (Leyens et al., 2000).

We presented participants with the list of six primary and six secondary emotions and asked them to rate how likely they felt it was the offender was generally capable of expressing those emotions, using a 5-point response scale (1 = extremely unlikely to 5 = extremely likely). We expected that those in the compassion-enhanced condition would have significantly higher scores than those in the standard condition, with respect to attributing secondary emotions to offenders.

Adequacy of sentence

Finally, we asked participants about the adequacy of the sentence imposed. Here, they responded on a 5-point scale whether the sentence was 1 = much too lenient to 5 = much too harsh. Here, we anticipated that participants in the compassion-enhanced condition would be more likely to view the sentence as being appropriate or too harsh.

Covariates

We also included several covariates. Specifically, we controlled for demographic factors, including gender (male = 1; female = 0); age (18–24 years = 1; other = 0); education (Bachelor’s degree or higher = 1; other = 0); and political affiliation (Liberal/National Party = 1; other = 0). Previous research suggests that men are more likely to support harsher penalties (e.g. Applegate et al., 2002), and punitiveness increases with age (Fitzgerald et al., 2018). Research also consistently shows that lower education is linked to more punitive attitudes (Spiranovic et al., 2012), as is political conservatism (Costelloe et al., 2009).

We also controlled for additional potential covariates, including participants’ CJS experiences – namely, having attended court as a witness or support person; having served on a jury; knowing someone who had been imprisoned; and having worked in the CJS (all yes = 1; no = 0). Research in this area is somewhat limited, but suggests, for example, that experiencing jury duty increases a person’s belief in the equity and justice of the legal system (Allen, 1977); individuals working in justice-related fields favour non-punitive interventions (Ortet-Fabregat & Pérez, 1992); and having an immediate family member or close friend charged with an offence makes one more likely to perceive the CJS as too harsh (Davila et al., 2011).

Experiment 1

The goal of Experiment 1 was to examine the effect that more compassionate sentencing remarks, when presented to participants in written form, have on the public’s compassion toward those who offend, spending preferences and punitiveness. We also examined the impact on participants’ views of whether offenders are similar to themselves, the capacity of offenders to feel various emotions and the adequacy of the sentence imposed.

Method

Participants

The sample comprised 216 members of the Australian public recruited by Qualtrics, on behalf of the research team, to complete an online survey. Survey responses were collected over 10 days in February 2020, until a minimum quota was reached. The inclusion criteria required individuals to be aged 18 or over and living in Australia. Participants received an incentive from Qualtrics, following completion of the survey.

Most participants (n = 113; 52.3%) were female. A large proportion were aged 55 years and over (n = 86; 39.8%) and had a Bachelor’s degree or higher as their highest level of educational achievement (n = 75; 35%). In terms of political preference, nearly two fifths of participants (n = 80; 37%) identified as supporters of the Australian Labor Party, 28% (n = 61) supported the Liberal/National Party (LNP), while 20% (n = 44) did not support any Australian political party. Smaller proportions of respondents supported the Australian Greens (n = 18; 8%), while 5% (n = 11) said they supported an unlisted political party, and 1% (n = 2) responded they did not know.

Design and manipulations

The survey in Experiment 1 was designed so that participants were randomly assigned by Qualtrics to one of the two written versions (i.e. standard or compassion-enhanced) of remarks imposing a sentence on an offender who had pleaded guilty to aggravated burglary, based on the actual remarks of Justice Elkaim in R v Wickes (2018). The original sentence was chosen partly because Justice Elkaim’s reasons were stated simply and intelligibly, and incorporated factual findings and observations relating to the suffering experienced by the offender and his victims. Consequently, the primary text of each judgment was mostly identical. No additional factual material about the offence or offender was added in the process of compassion enhancement. Notwithstanding this, the compassion-enhanced version was significantly longer (1570 vs. 1061 words), largely because it provided additional explanation about the sentencing process and relevance of factual material contained in the sentence. The standard and compassion-enhanced remarks are provided in full in Supplementary Appendices C and D, respectively. A table outlining the key differences between the versions is provided in Supplementary Appendix E.

In the standard version, the person being sentenced was consistently referred to as ‘the offender’. In contrast, in the compassion-enhanced version, the offender is consistently addressed both by his full name and using the second-person pronoun (i.e. ‘you’). The purpose was to convey, through relational, non-objectifying language, the fact that the judge was speaking directly to, and remaining turned towards, the person being sentenced. It was hypothesised that this change had the capacity to highlight and support the recognition by members of the public of all definitional elements of compassion.

All material added to the compassion-enhanced sentencing remarks was intended to increase recognisable features of compassion being shown towards the offender by the sentencing judge. An example of this is the inclusion of two short paragraphs at the beginning of the remarks, to explain to the offender, James Wendel, what must be considered in the sentencing process. These were inspired, in part, by the sentencing remarks of Chief Judge Kidd in DPP v Pell (2019). Although Pell was subsequently acquitted on appeal by the High Court, the sentencing remarks remain available on various publicly accessible websites. The additions are also examples of relational language:

James Wendel, I sentence you today on the basis that you are a member of our community, just like any other.

I must consider the circumstance of the offence you committed, including the harm that you caused. I must also consider your personal circumstances. This is important because I must impose a sentence that is just and appropriate. The sentence must also be designed to achieve the best possible outcome for you and the community.

The purpose here was to display, from the outset, universality in the form of common humanity, a recognition of suffering and a motivation to act to alleviate that suffering. We hypothesised that these additions would also recognisably convey empathy and connection, as well as a willingness to tolerate distress.

At other points, additional material was added, to explain the relevance of evidence before the court. This provided an opportunity to explicitly convey features of compassion. For example, the standard version included a powerful and compassionate quote from the victim impact statement of the mother of a child held at knifepoint during the burglary:

Accept your punishment. Get help, so you can become someone who doesn’t need drugs. Make a positive contribution to society and live a happy, long life, surrounded by love and support. You’re only young and you’ll still have your whole life ahead of you. This incident does not define you. People change.

In the standard version, the quote was followed by the words ‘I can only hope that the offender takes note of this’. In the enhanced version, we sought to convey, even more strongly and personally, the judge’s support for the compassionate elements of this statement, highlighting the recognition of suffering, an understanding of its universality as a human experience and a commitment to its alleviation, by conveying empathy, through an understanding of the whole person, not defined by his offending (Winick & Wexler, 2003):

James Wendel, I hope that you can listen to what she says. It is important that you face what you have done and the harm that you have caused. But it is also important that you understand that what you have done does not define you. I too hope that you can find a way to make a positive contribution to society and live a happy, long life, surrounded by love and support.

At various points in the compassion-enhanced remarks, material was added to more directly demonstrate the judge’s focus on recognising and understanding the causes and conditions of the offender’s suffering, related to his drug use and mental health, which also had a clear connection with his offending. For example, after a quote from a forensic psychiatrist that provided evidence of the connection between the offence and a ‘substance-induced psychotic disorder’, which appeared in both versions, the following was added to the compassion-enhanced version:

This is important. It provides some explanation for the offence. It makes clear that it is essential for you to get treatment and support to address your drug and alcohol dependency issues.

Finally, in recognition that the court was required to decide whether to imprison the offender, and for how long, additional material was added to reflect the judge’s understanding of the likelihood that prison might have negative consequences for the offender and the community, thereby inflicting further suffering, without acting to alleviate this. The standard remarks include the following statement: ‘I must keep in mind the principle that a person should not be sent to prison other than as a last resort’. Again, drawing inspiration from the remarks in DPP v Pell (2019), the following was added to the compassion-enhanced version:

I must decide whether to send you to prison for this offence, and, if so, for how long. You have not served a term of imprisonment before. Prison may be very difficult for you. It may have negative consequences for you and your rehabilitation. This would be greatly to the community’s disadvantage.

After reading either the standard or enhanced remarks, participants read a short statement by the judge, describing the sentence imposed, which was identical across the two conditions.

To ensure all participants had read and adequately understood the sentencing remarks, they were given seven statements regarding the contents of the remarks and were asked whether these statements were true or false, yielding a comprehension score out of 7. Most participants (79.4%; n = 216) gave at least four correct answers. The remaining participants (20.6%; n = 56), who answered three or fewer questions correctly, were excluded from the analysis, given concerns they had not read and/or understood the remarks. This left an analytical sample of 216 participants. Of these, 107 (49.5%) received the compassionate remarks, and 109 (50.5%) received the standard remarks.

After reading the remarks, participants were presented with a series of statements. These included several questions designed to evaluate the effectiveness of the intended manipulated differences between the two conditions (e.g. ‘did the judge recognise the suffering of the offender?’). Further, both before and after reading the remarks, participants responded to five questions on whether compassion should be shown to offenders, which were the five items derived from the Strauss et al. (2016) study. By repeating these questions at two points in the survey, we could establish a baseline measure for participants’ compassion and assess the impact of reading the sentencing remarks in each condition on the importance of compassion toward offenders. Participants were also asked to complete the criminal justice budget allocation task and were given questions to measure their support for punitive responses to offending. Finally, we asked participants their views on whether the offender was similar or not to them, the capacity of offenders to feel various emotions (the infrahumanisation measure) and their views on the adequacy of the sentence imposed.

Results and discussion

Manipulation check

We conducted a series of independent t tests to ascertain whether there was a significant difference between the two groups on the manipulation check questions (see Table 1).

The t-test results revealed a significant difference, t(204) = 2.61, p = .010, η² = .03, in participants’ mean scores, when asked whether ‘the judge recognised the suffering of the offender’. The participants in the compassion-enhanced condition had higher mean scores (M = 3.92, SD = 0.95) than participants in the standard condition (M = 3.59, SD = 0.89), indicating that members of the former group felt the judge had shown a greater recognition of the offender’s suffering.

To explore these results further, we combined the five individual items, the indicators of compassion derived from Strauss et al. (2016), into a Manipulation Check Scale, where higher scores indicated a participant was more likely to recognise the compassionate components of the sentencing remarks. The construction of this scale was supported by a principal components analysis, resulting in a one-factor solution. A reliability analysis showed the scale had a Cronbach’s alpha value of .87. A t-test analysis on the scale scores that resulted from combining across the five manipulation check items found a significant difference between participants’ mean scores (standard group M = 3.45, SD = 0.72; compassionate group M = 3.66, SD = 0.81), t(204) = 2.02, p = .045, η² = .02, again confirming that those in the compassion-enhanced condition were more likely to recognise that the judge was being more compassionate than in the standard condition. The effect size of this difference was, however, relatively small.

Importance of showing compassion toward offenders

Next, to evaluate whether the manipulated differences in the remarks had any impact on participants’ general feelings about the importance of being compassionate toward offenders, we examined participants’ responses to the five compassion questions (which were asked both before and after reading the remarks). To ascertain what effect, if any, the compassion-enhanced remarks had on participants’ beliefs that it is important for judges to demonstrate compassion towards offenders when sentencing, we calculated participants’ mean difference in Importance of Compassion scores (by subtracting the pre-remark Importance of Compassion score from the post-remark Importance of Compassion score). A series of independent t tests on participants’ mean difference scores for the five individual items (for all scores, see Table 2) found no significant pre–post differences in participants’ scores, suggesting that the version read had no effect on belief in the importance of demonstrating compassion toward offenders.

To explore this further, we combined the questions into a five-item Importance of Compassion Scale, which was supported by a principal components analysis, showing a one-factor solution. A reliability analysis showed a Cronbach’s alpha value of .58, just less than the traditionally accepted cut-off value for scale construction (Bonett & Wright, 2015).

The t-test analysis confirmed there was no significant difference between the groups on this scale (compassion-enhanced group, M = 0.09, SD = 0.49; standard group, M = 0.09, SD = 0.50), t(187) = 0.10, p = .917, η² = .00. As a final step, we tested an analysis of covariance (ANCOVA) model on the Compassion Scale, to test the alternative hypothesis that the mean Importance of Compassion difference scores were equal once several relevant covariates were accounted for. As set out in Table 3, there remained a non-significant difference between the two groups, after controlling for the covariates.

Table 3.

Experiment 1: ANCOVA results for participants’ compassion scale scores.

	df	F	MSE	Effect size(${\mathrm{\eta }}_{\mathrm{p}}^2$)
Male	1	0.59	0.15	.003
Age 18–24 years	1	0.00	0.00	.000
Bachelor’s degree or higher	1	0.51	0.13	.003
Political party – Liberal National Party	1	0.62	0.16	.003
Attended court as witness/support person	1	0.16	0.04	.001
Served on jury	1	0.36	0.09	.002
Known someone imprisoned	1	0.23	0.06	.001
Worked in CJS	1	0.43	0.11	.002
Version of remarks	1	0.00	2.71	.000

Note: ANCOVA = analysis of covariance; CJS = criminal justice system; compassion-enhanced remarks n = 95; standard remarks n = 94.

Budget allocation task

As the final step, we analysed participants’ criminal justice spending preferences. As shown in Table 4, there were differences in the rank order of and percentage allocation to the five criminal justice initiatives between the two groups. Specifically, the highest spending priority for participants in the compassion-enhanced condition had a rehabilitative focus, funding more alcohol and drug treatment programmes (25.48%), followed by more programmes aimed at the welfare of victims (23.63%). The third highest preference was for more police on the street (21.98%), followed by more mental health treatment programmes (21.20%), and building more prisons (7.71%).

Table 4.

Experiment 1: Participants’ mean spending preferences for various rehabilitative and punitive crime control measures, by standard and compassionate groups.

	Group				Effect size(η2)
	Compassion-enhanced(n = 90)		Standard(n = 89)
	Rank	Mean spending(%)	Rank	Mean spending(%)
More prisons*	5	7.71	5	11.94	.037
More drug/alcohol treatment programmes for offenders*	1	25.48	3	20.89	.033
More police on the street	3	21.98	4	20.20	.003
More programmes aimed at the welfare of victims	2	23.63	1	24.63	.002
More mental health treatment programmes for offenders	4	21.20	2	22.33	.002

*p < .05.

For participants in the standard condition, victim welfare (the only neutral option available to participants) took priority, with 24.63% of funds allocated here. This was followed by more mental health treatment (22.33%), drug and alcohol treatment (20.89%) and police (20.20%). This group also ranked spending on more prisons as their least preferred option (11.94%).

Next, to determine whether participants’ spending allocations on any of the five items were significantly different, we conducted a series of t tests on the proportion of funds allocated by participants in each condition to each of the five spending options. As shown in Table 4, t tests showed there was a statistically significant difference, t(160) = –2.60, p = .010, η² = .04, in spending preferences, with respect to ‘building more prisons’. A significant difference, t(177) = –2.45, p = .015, η² = .03, was also found with respect to participants’ preferred spending on drug and alcohol treatment programmes.

As a final step in this analysis, we conducted two one-way ANCOVA analyses, to determine whether there was an overall statistically significant difference in spending on more prisons (Table 5) and drug and alcohol treatment (Table 6), once mean values were adjusted for four demographic covariates (gender, age, education level and political affiliation) and four covariates capturing participants’ experiences of the criminal justice system. Before conducting these analyses, we carried out assumptions testing, to ensure the reliability of the ANCOVA results. All necessary assumptions for ANCOVA were satisfied.

Table 5.

Experiment 1: ANCOVA results for participants’ preferred spending on ‘more prisons’.

	df	F	MSE	Effect size(${\mathrm{\eta }}_{\mathrm{p}}^2$)
Male	1	6.16*	715.79	.035
Age 18–24 years	1	1.97	229.25	.012
Bachelor’s degree	1	1.02	118.33	.006
LNP	1	0.41	48.24	.002
Attended court	1	1.49	173.88	.009
Jury service	1	0.33	37.94	.002
Known someone imprisoned	1	0.46	54.13	.003
Worked in CJS	1	0.02	2.28	.000
Version of remarks	1	9.27**	1077.74	.052

Note: ANCOVA = analysis of covariance; LNP = Liberal/National Party; CJS = criminal justice system; compassion-enhanced remarks n = 90; standard remarks n = 89.

*p < .05. **p < .01.

Table 6.

Experiment 1: ANCOVA results for participants’ preferred spending on ‘more drug and alcohol treatment programmes for offenders’.

	df	F	MSE	Effect size(${\mathrm{\eta }}_{\mathrm{p}}^2$)
Male	1	4.39*	656.89	.025
Age 18–24 years	1	0.56	83.84	.003
Bachelor’s degree	1	5.67*	847.03	.032
LNP	1	0.19	28.03	.001
Attended court	1	0.15	22.86	.001
Jury service	1	0.08	12.65	.001
Known someone imprisoned	1	1.22	182.36	.007
Worked in CJS	1	5.14*	768.02	.030
Version of remarks	1	7.77**	1162.10	.044

Note: ANCOVA = analysis of covariance; LNP = Liberal/National Party; CJS = criminal justice system; compassion-enhanced remarks n = 90; standard remarks n = 89.

*p < .05. **p < .01.

In the first model (for spending on more prisons), the ANCOVA analysis showed that a significant difference, F(1, 169) = 9.27, p = .003, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .05, between the two groups remained, after controlling for the eight covariates. The adjusted mean for participants in the standard condition increased to 12.36%, but decreased to 7.30% for those in the compassion-enhanced condition. In this model, 4.8% of the variation in the outcome variable means was explained by the grouping variable and covariates.

In the second model (for spending on more drug and alcohol treatment), the ANCOVA analysis also showed that a significant difference, F(1, 169) = 7.77, p = .006, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .04, remained between, after accounting for the eight covariates. Here, the adjusted mean for participants in the standard condition reduced to 20.56%, while it increased to 25.8% for those in the compassion-enhanced condition. In this model, 7.5% of the variation in mean spending was explained by the grouping variable and covariates.

Punitiveness toward offenders

To evaluate whether the manipulated differences had any impact on punitiveness, we examined participants’ responses to the seven questions measuring their support for the harsher treatment of offenders. These questions were asked before and after they read the sentencing remarks, with participants’ mean difference in punitiveness scores calculated. The calculated t tests on each of the seven items found no significant difference between the groups, contrary to our hypothesis.

Group membership

We combined the group membership items into a two-item scale, which was supported by a principal components factor analysis and reliability analysis (Cronbach’s alpha = .84). On this scale, higher scores indicated a participant was more likely to see the offender as someone like themselves. A t-test analysis found no significant difference in scale scores between participants in the two conditions.

Attribution of emotions

Next, we examined the six primary and six secondary emotions that participants were asked how capable the offender was of experiencing. Contrary to our hypothesis, t-test analyses on each individual emotion found no significant difference between participants in the compassion-enhanced condition, compared to the standard condition. There were also no significant differences between conditions using the composite score approach as has been used by infrahumanisation researchers in the past (Gaunt, 2009; Viki & Calitri, 2008).

Adequacy of sentence

Finally, we examined whether participants’ mean scores differed, when asked their views on the adequacy of the sentence imposed. Contrary to predictions, no significant difference was found here, with both groups appearing to view the sentence as being between ‘too lenient’ and ‘about right’.

Conclusion

We tested the effect of using compassion-enhanced sentencing remarks on seven possible outcomes. Our analysis revealed several key findings.

Firstly, our results show that we were successful in altering the sentencing remarks in the case of R v Wickes (2018), to introduce more detectable and apparent features of compassion toward the offender being sentenced. Specifically, participants in the compassion-enhanced condition were significantly more likely to agree that the judge had recognised the suffering of the person being sentenced. Furthermore, when all of the five compassion indices were combined into a single scale, participants in the compassion-enhanced condition also had significantly higher scores than participants in the standard condition.

Second, our results provide empirical support for Strauss et al.’s (2016) conceptualisation of compassion as having five relevant dimensions. For this experiment, we adapted these five dimensions to reflect a criminal justice context and combined the items to create a Manipulation Check Scale and an Importance of Compassion Scale. A factor analysis confirmed there was a single underlying factor or concept to both of those scales. Thus, these scales derived from Strauss et al.’s conceptualisation of compassion appear to be a conceptually valid scale and should assist researchers in future research.

However, despite the soundness of the Manipulation Check Scale and Importance of Compassion Scale, and contrary to our hypothesis, our results found no significant difference between participants’ scores on the Importance of Compassion Scale across the two conditions. Specifically, while our results showed that our intended manipulations of the sentencing remarks were successful (with participants recognising the features of compassion we had emphasised in the compassion-enhanced condition), we found no significant difference between the groups’ Importance of Compassion scores. Put simply, while participants in the compassion-enhanced condition did recognise the more compassionate elements of the sentencing remarks, this did not significantly alter their views on the compassion that ought to be shown to those who offend. At this point we wondered whether delivery mode used (reading judgments versus hearing orally delivered sentencing remarks) may have an impact on importance of compassion scores between conditions, hence the thought to design Experiment 2 as reported below.

Our third result showed, however, that the presence of compassion in the sentencing remarks did have an impact on participants’ criminal justice spending preferences. Specifically, participants in the compassion-enhanced condition allocated more government funding on a rehabilitative option as their first preference – more drug and alcohol treatment programmes – spending just over one quarter of their available funds (25.48%) on this option. In contrast, participants in the standard condition spent significantly less on this option (20.89%), ranking it as their third spending preference. Instead, standard condition participants allocated most of their funding (24.63%) to the neutral option available – more programmes aimed at victim welfare.

Interestingly, spending on more prisons was the least preferred option for both groups, with those in the standard condition allocating 11.94% and compassion-enhanced participants allocating just 7.71%. This is despite the political rhetoric suggesting that the public supports widespread criminal justice spending on prisons. Consistent with our hypothesis, though, a statistically significant difference between the groups was observed here, with those in the standard condition allocating a greater amount of taxpayer money to this option.

Finally, this experiment showed that, contrary to our hypotheses, the more apparent features of compassion in the sentencing remarks did not impact participants’ views on how harshly offenders ought to be treated by the CJS, their identification of the offender as someone like them, or the types of emotions they felt an offender could experience. It also had no impact on their views of the adequacy of the sentence imposed.

Experiment 2

The goal of Experiment 2 was to examine what effect compassion-enhanced sentencing remarks, when presented to participants in audio-visual (i.e. video) form, have on the public’s compassion for those being sentenced, spending preferences and desires for the harsher treatment of offenders (and how these effects differ from the presentation of sentencing remarks in written form). Further, we examined the impact of the altered remarks on participants’ views about whether offenders are similar to themselves, their capacity to experience a variety of emotions and the adequacy of the sentence given.

Method

Participants

The analytical sample for Experiment 2 comprised 232 members of the Australian public, who were again recruited and compensated by Qualtrics, using an online survey with the same eligibility criteria. Survey responses were collected between July and October 2020. Individuals who failed to correctly answer at least four comprehension questions about the survey were automatically screened out by Qualtrics. Participants had a similar demographic profile to those in Experiment 1; most participants were female (60%; n = 139), and a large proportion were aged 55 years and over (31%; n = 71), and had at least a Bachelor’s degree (44%; n = 103). For political preference, participants were fairly evenly split between supporting the Australian Labor Party (33%; n = 77) and LNP (35%; n = 81), while 14% (n = 32) said they did not support any Australian political party. A smaller number of respondents supported the Australian Greens (10%; n = 24), an unlisted political party (5%; n = 11) or ‘did not know’ (3%, n = 7).

Design and manipulations

Participants were randomly assigned by Qualtrics to one of the two video-recorded versions (i.e. standard or compassion-enhanced) of the sentencing remarks used in Experiment 1.

The standard videoed version of the remarks showed a male member of the research team acting as the sentencing judge (see Figure 1). He wore judicial robes and was sitting at a court bench (in a moot courtroom at the Australian National University), facing the body of the court and not turned towards the offender. The video was framed to show the ‘judge’ in a seated position, from the mid-section upward. The ‘judge’ ensured that his voice remained monotone, with no noticeable emotion. He limited his eye contact with the position where the offender would be seated and instead addressed the broader court in his delivery of the remarks. His facial expressions were neutral throughout. The sentencing remarks read by the judge were identical to the written remarks in the standard condition in Experiment 1, with the judge referring only to ‘the offender’. The videoed version lasted six-and-a-half minutes.

Figure 1.

Experiment 2: Still image from delivery of ‘standard’ sentencing remarks. To view this figure in colour, please visit the online version of this Journal.

In the compassion-enhanced version of the remarks, the judge is portrayed by the same person, and the video is framed to show him seated at the bench from his mid-section upward (see Figure 2). In this condition, the judge is turned slightly to his left, facing where the offender would be seated. Following the recording of the standard remarks in the first condition, the actor portraying the judge took a short break to complete a compassion meditation exercise, inspired by previous research (e.g. Falconer et al., 2019; Klimecki et al., 2013). This exercise involved evoking an image of James, the person being sentenced, understood from information contained in the sentence and past experience with offenders who have faced substance dependency and mental illness, and then silently repeating compassionate phrases towards James, connecting with a genuine aspiration or wish for James’s suffering to end, for example, ‘may you be free from suffering’ or ‘may you be free from addiction’. The purpose was to generate, through an imaginative exercise that could be replicated by a judge, prior to delivering a sentence in a real case, a felt sense or experience of compassion and connection, which might convey itself (and continue to be felt), during the delivery of sentencing remarks, through facial expression, voice tone and body language. The instructions for this compassion meditation primer can be found at Supplementary Appendix F. The ‘judge’ sought to portray more warmth and emotion through his voice (Cordaro et al., 2016; Simon-Thomas et al., 2009) and have increased eye contact and a steady gaze (Baránková et al., 2019) toward the offender, during his delivery of the sentencing remarks.

Figure 2.

Experiment 2: Still image from delivery of ‘compassion-enhanced’ sentencing remarks. To view this figure in colour, please visit the online version of this Journal.

The sentencing remarks in this condition were identical to those in the compassion-enhanced version in Experiment 1, whereby the person being sentenced is addressed by name, and additional comments are made to recognise the impact of the events on him and to highlight his attempts and prospects of rehabilitation. This version of the remarks lasted for 10.5 minutes.

Out of the total of 232 survey responses, 124 participants (53%) were randomly allocated to the standard version of the videoed sentencing remarks, and 108 participants (47%) to the compassion-enhanced remarks.

Results and discussion

Manipulation check

The survey included the same five manipulation check questions as those in Experiment 1, to determine whether participants in the compassion-enhanced condition did in fact detect more compassionate features than participants in the standard condition (see Table 7). Analysis on the individual items via separate t tests showed a significant difference, t(220) = 2.09, p = .038, η² = .02, in participants’ mean scores for whether ‘the judge was motivated to alleviate the suffering of the offender’ (albeit a relatively small effect size). Accordingly, compassionate group participants had higher mean scores (M = 3.75, SD = 0.92) than standard group participants (M = 3.46, SD = 1.10), indicating a greater level of agreement with this statement.

The five manipulation check questions were also combined into a Manipulation Check Scale (supported by a one-factor solution on a principal components analysis and with the scale having a Cronbach’s alpha = .87), where higher scores indicate a person is more likely to recognise the compassionate components of the sentencing remarks. Unlike in Experiment 1, a t-test analysis found no significant difference in participants’ mean scores on this scale.

Compassion toward offenders

As in Experiment 1, we evaluated the effect of the sentencing remark condition on participants’ belief that it is important to express compassion toward offenders. Participants responded to five statements regarding their compassion for offenders twice during the survey, before and after they viewed the sentencing remarks (see Table 8). The difference between their pre- and post-remark Importance of Compassion scores were calculated, and a series of independent t tests were conducted on the mean difference scores, to test our hypothesis that a larger pre–post change would be observed amongst those in the compassion-enhanced condition than the standard condition.

As shown in Table 8, significant differences between the two groups were identified on two of the Importance of Compassion items. Firstly, a significant difference, t(198) = 2.87, p = .005, η² = .04, in participants’ mean scores was found in relation to whether ‘we should recognise the suffering of offenders’. Specifically, those in the compassion-enhanced group had larger mean difference scores (M = 0.47, SD = 0.87) than standard group participants (M = 0.13, SD = 0.85), indicating that viewing the compassionate sentencing remarks had a greater impact on their belief in the importance of demonstrating compassion toward offenders. A significant difference, t(167) = 2.02, p = .045, η² = .02, was also found when asked whether ‘we should feel emotional responses toward offenders who suffer’. There, compassionate group participants again had larger mean difference scores (M = 0.32, SD = 0.98) than standard group participants (M = 0.08, SD = 0.67).

The five individual compassion items were then combined into an Importance of Compassion Scale. The construction of this scale was supported by a principal components factor analysis, showing a one-factor solution, although a reliability analysis revealed a Cronbach’s alpha of .67, just less than the traditionally accepted value of .70. A t-test analysis on the Compassion Scale found a statistically significant difference between the groups, t(173) = –3.09, p = .002, η² = .05. Specifically, participants in the compassion-enhanced condition had significantly higher mean scores (M = 0.38, SD = 0.62) than those in the standard condition (M = 0.14, SD = 0.44), indicating that that version of the videoed sentencing remarks had a greater effect on increasing participants’ beliefs in the importance of compassion for offenders.

We then ran an ANCOVA analysis on the Importance of Compassion Scale, to test the null hypothesis that the two groups’ mean scale scores were equal, after controlling for several demographic and criminal justice experience covariates (see Table 9). This analysis showed that, even when controlling for the eight covariates, a significant difference between the two groups remained, F(1, 190) = 9.10, p = .003, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .05. The adjusted mean difference score for participants in the standard condition was .15 versus .38 for those in the compassion-enhanced condition. In this model, 3.5% of the variation in the outcome variable was explained by the grouping variable and covariates.

Table 9.

Experiment 2: ANCOVA results for participants’ compassion scale scores.

	df	F	MSE	Effect size(${\mathrm{\eta }}_{\mathrm{p}}^2$)
Male	1	3.33	0.96	.017
Age 18–24 years	1	0.77	0.22	.004
Bachelor’s degree	1	1.14	0.33	.006
LNP	1	0.23	0.07	.001
Attended court	1	0.03	0.01	.000
Jury service	1	0.01	0.00	.000
Known someone imprisoned	1	0.54	0.16	.003
Worked in CJS	1	0.29	0.08	.002
Version of remarks	1	9.10**	2.63	.046

Note: ANCOVA = analysis of covariance; LNP = Liberal/National Party; CJS = criminal justice system; compassion-enhanced remarks n = 97; standard remarks n = 103.

**p < .01.

Budget allocation task

The criminal justice budget allocation task from Experiment 1 was also presented to participants in Experiment 2. As shown in Table 10, a t-test analysis showed there were statistically significant differences in participants’ spending preferences. Specifically, participants differed in their spending preferences for providing ‘more mental health treatment programmes for offenders’, with those in the compassion-enhanced condition allocating significantly more funding to this option (standard group, M = 18.55, SD = 12.92; compassion-enhanced group, M = 23.26, SD = 13.69), t(203) = 2.53, p = .012, η² = .03.

Table 10.

Experiment 2 Participants' mean spending preferences for various rehabilitative and punitive crime control measures, by standard and compassionate groups.

	Group				Effect size(η2)
	Compassion-enhanced		Standard
	Rank	Mean spending(%)	Rank	Mean spending(%)
More prisons	5	10.51	5	14.47	.017
More drug and alcohol treatment programmes for offenders	3	21.55	1	22.48	. 001
More police on the street	4	20.34	2	22.28	.003
More programmes aimed at the welfare of victims	1	24.33	3	22.21	.007
More mental health treatment programmes for offenders*	2	23.26	4	18.55	.030

Note: *p < .05; Compassion-enhanced group n = 93; standard group n = 112.

To further examine this finding, we used ANCOVA to explore the effect, if any, of several covariates (see Table 11). That model shows that a significant difference remained in mean spending preferences on mental health treatment between the two groups, even after controlling for the eight demographic and justice-related variables, F(1, 195) = 8.05, p = .005, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .04. The adjusted mean for standard group participants decreased to 18.30, while it increased to 23.56 for those in the compassion-enhanced group. In this model, 6% of the variation in the outcome variable means is explained by the grouping variable and covariates.

Table 11.

ANCOVA results for participants’ preferred spending on ‘more mental health treatment programmes for offenders’.

	df	F	MSE	p	Effect size(${\mathrm{\eta }}_{\mathrm{p}}^2$)
Male	1	1.51	257.16		.008
Age 18–24	1	1.74	295.46		.009
Bachelor’s degree	1	0.69	117.35		.004
LNP	1	5.38*	914.27		.027
Attended court	1	1.50	254.62		.008
Jury service	1	0.57	97.27		.003
Known someone imprisoned	1	0.58	99.15		.003
Worked in CJS	1	0.86	145.80		.004
Version of remarks	1	8.05**	1368.50		.040

Note: ANCOVA = analysis of covariance; LNP = Liberal/National Party; CJS = criminal justice system; compassion-enhanced remarks n = 93; standard remarks n = 112.

*p < .05. **p < .01.

Punitiveness toward offenders

We then examined participants’ responses to the seven questions relating to the harsher treatment of offenders. We calculated participants’ mean difference punitiveness scores and conducted t tests on each individual item. Contrary to our hypothesis, no significant difference between the groups was found.

Group membership

We combined the group membership items into a two-item scale, which was supported by a principal components factor analysis, showing a one-factor solution and a Cronbach’s alpha of .70. Again, mirroring the results of Experiment 1, and contrary to our hypothesis, a t-test analysis found no significant difference between the two conditions.

Attribution of emotions

We then examined the primary and secondary emotions, asking participants how likely it was an offender could experience these emotions. As in Experiment 1, and again contrary to our hypothesis, t-test analyses on each individual emotion found no significant difference between participants across the two conditions. Separate t tests on a primary emotion and secondary emotion composite scores also revealed no significant difference between the groups.

Adequacy of sentence

Finally, we examined participants’ mean scores about their views on the adequacy of the sentence imposed. Contrary to our hypothesis, but in line with the results of Experiment 1, no significant difference was found, with both groups again viewing the sentence as being between ‘too lenient’ and ‘about right’.

Conclusion

The goal of Experiment 2 was to test the effect of enhancing compassion in videoed sentencing remarks, to ascertain whether the presence of audio-visual compassion cues, shown by the judge toward the offender, impacted participants’ compassion toward offenders or spending preferences. Our analysis revealed several notable results.

To begin, while our manipulations between the two conditions were successfully recognised by participants in Experiment 1, manipulation check results for Experiment 2 were not as clear. While participants in Experiment 2 in the compassion-enhanced condition were significantly more likely to agree that ‘the judge was motivated to alleviate the suffering of the offender’, no significant difference between the groups was found on any other individual item, nor on a scale combining participants’ scores across the five manipulation check questions. This result was promising, in showing we were somewhat successful in manipulating compassionate delivery between the two conditions, so that one contained more noticeable features of compassion in the judge’s body posture, tone of voice and the substantive comments made within the audio-visual sentencing remarks. However, it also suggests that the features of compassion we had intended were not as strongly identified by participants as we had hoped, and these features may need to be emphasised further or in different ways in future research.

Despite these weaker results on our manipulation check items, Experiment 2 revealed some interesting effects of the two conditions on participants’ beliefs about the importance of demonstrating compassion toward offenders. Specifically, participants in the compassion-enhanced condition were significantly more likely to agree that it is important to recognise offenders’ suffering and that we should feel emotional responses toward offenders who suffer. Moreover, participants in that condition also had significantly higher scores on the Importance of Compassion Scale, with these results holding even after we had controlled for several relevant demographic factors. This suggests that viewing the delivery of recognisably more compassionate oral sentencing remarks can impact how the public feels about people being sentenced and, importantly, has the capacity to increase public willingness to recognise their suffering. In this way, Experiment 2 also shows how compassionate sentencing has the capacity to change how we think and feel about those who offend.

In addition, this experiment showed that the presence of compassion can impact the public’s preferences on the allocation of funding to criminal justice responses. Specifically, participants in the compassion-enhanced condition were significantly more likely to allocate funding to the provision of more mental health treatment programmes, spending 23.26% of their funding here, compared to 18.55% for those in the standard condition. Unexpectedly, though, participants who viewed the compassion-enhanced remarks allocated the largest proportion of funding (24.33%) to the neutral option – providing more programmes aimed at the welfare of victims – while participants in the standard condition allocated their largest proportion of funds (22.48%) to a rehabilitative option – more drug and alcohol treatment programmes. As in Experiment 1, neither group favoured large additional spending on more prisons.

Finally, several of our results mirrored those of Experiment 2. Specifically, contrary to our expectations, the delivery of more compassionate sentencing remarks was not found to significantly impact participants’ views on how harshly offenders ought to be punished, their positioning of the offender as someone like themselves, or their attribution of specific emotions as capable of being experienced by offenders. Nor was any impact found with respect to participants’ views on the adequacy (or not) of the sentence given.

General conclusion

Taken together, the results of these two experiments are encouraging for researchers and judicial officers alike, as they provide some evidence that Strauss et al.’s (2016) five components of compassion can be incorporated into sentencing remarks, in a way that is detectable by the audience of those remarks. However, more research is required on how such components can and should be used as the basis for improving the compassionate nature of judicial communication, as well as shaping communication between other legal actors or authority figures and offenders in a criminal justice context. Further links between compassionate delivery in both written and oral judgments, and the impact of those features on public attitudes towards offenders and sentencing should be investigated.

Interestingly, when sentencing remarks were viewed by participants, as they would be in court, there were some even clearer effects seen for the impact of compassionate delivery on perceptions that it is important to be compassionate whilst sentencing. Across both experiments, there was little interest in spending on prisons, but instructive insights emerged into and shifts in how participants in the compassionate sentencing conditions, relative to the standard sentencing conditions, would like to see justice budgets spent. The results suggest that compassionate sentencing shifted participants’ support towards more spending on drug and alcohol treatment (in Experiment 1) and mental health treatment (in Experiment 2).

Interestingly, the infrahumanisation and other measures aimed at investigating the intergroup and moral community divisions that sentence delivery choices can construct between offenders and non-offending members of the public were not, at least in these experiments, clearly influenced by manipulating the compassionate features of the sentencing remarks and their delivery in the ways we did. It should also be noted that the observed effects had relatively small effect sizes. Our results must also be interpreted in light of the relatively small sample sizes in each experiment, although the study does not appear underpowered, and the fact that these samples were not necessarily representative of the broader Australian population.

In Experiment 2, a further limitation is that only one judicial officer delivered the remarks, and future research should test whether the manipulated features of compassionate delivery can generalise across different judicial officers, with a diversity of judicial demographics and characteristics tested. Future studies might also examine the impact of gender, with research indicating that female (but not male) respondents tend to be less punitive toward female offenders (Applegate et al., 2002). Thus, both the gender of the offender and the respondent may influence the level of compassion shown. Furthermore, these experiments did not test how the sentenced person themselves experienced the impact of compassion-enhanced sentencing. To test some of those dynamics, interviews or experiments with imprisoned and other sentenced persons may be revealing, to determine how sentenced persons are affected by the presence or absence of compassionate features of sentencing remarks and delivery. Finally, these experiments focused on the compassion felt by respondents toward offenders and did not simultaneously measure respondents’ feelings of compassion for victims. Future research might explore how respondents balance feelings of compassion for both offenders and victims, when provided with compassion-enhanced sentencing remarks. These initial results, however, suggest that it may be possible to reduce the public’s punitive reactions to offending, when the judicial officer turns towards the offender and sentences them in a way consistent with the calls made by the therapeutic jurisprudence literature and those recognising the important role compassion has in judicial work. That alone is encouragement to judicial officers keen to use compassionate features in the way they write and/or orally deliver remarks to those they sentence, when those offenders are considered to be fellow humans deserving of compassion, as well as justice.

Ethical standards

Declaration of conflicts of interest

Anthony Hopkins has declared no conflicts of interest

Shannon Dodd has declared no conflicts of interest

Mark Nolan has declared no conflicts of interest

Lorana Bartels has declared no conflicts of interest

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the Australian National University Human Research Ethics Committee Approval 2019/579 and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed consent

Informed consent was obtained from all individual participants included in the study

Funding Statement

This work was supported by an ANU College of Law Research Grant.

Supplemental data

Supplemental data for this article can be accessed here.