Miller-Lewis LR, Lewis TW, Tieman J, Rawlings D, Parker D, Sanderson CR (2021) Words describing feelings about death: A comparison of sentiment for self and others and changes over time. PLoS ONE 16(1): e0242848, Jan 6 2021. https://doi.org/10.1371/journal.pone.0242848
Rolf Degen's take: https://twitter.com/DegenRolf/status/1347045740305133572
Abstract: Understanding public attitudes towards death is needed to inform health policies to foster community death awareness and preparedness. Linguistic sentiment analysis of how people describe their feelings about death can add to knowledge gained from traditional self-reports. This study provided the first description of emotive attitudes expressed towards death utilising textual sentiment analysis for the dimensions of valence, arousal and dominance. A linguistic lexicon of sentiment norms was applied to activities conducted in an online course for the general-public designed to generate discussion about death. We analysed the sentiment of words people chose to describe feelings about death, for themselves, for perceptions of the feelings of ‘others’, and for longitudinal changes over the time-period of exposure to a course about death (n = 1491). The results demonstrated that sadness pervades affective responses to death, and that inevitability, peace, and fear were also frequent reactions. However, words chosen to represent perceptions of others’ feelings towards death suggested that participants perceived others as feeling more negative about death than they do themselves. Analysis of valence, arousal and dominance dimensions of sentiment pre-to-post course participation demonstrated that participants chose significantly happier (more positive) valence words, less arousing (calmer) words, and more dominant (in-control) words to express their feelings about death by the course end. This suggests that the course may have been helpful in participants becoming more emotionally accepting in their feelings and attitude towards death. Furthermore, the change over time appeared greater for younger participants, who showed more increase in the dominance (power/control) and pleasantness (valence) in words chosen at course completion. Sentiment analysis of words to describe death usefully extended our understanding of community death attitudes and emotions. Future application of sentiment analysis to other related areas of health policy interest such as attitudes towards Advance Care Planning and palliative care may prove fruitful.
Discussion
Summary of findings
Understanding public attitudes and feelings towards death is needed in order to inform health policies [3]. This study provided the first description of emotive attitudes of people asked to consider death utilising sentiment analysis. Applying multidisciplinary methodologies to a novel activity gave a deeper understanding of affective death attitudes in a community sample. We analysed the emotional sentiment of words people chose to describe feelings about death, for themselves, for perceptions of others, and for changes over the time-period during exposure to a course about death. The results demonstrated that text-sentiment analyses can provide a meaningful approach to death attitude research, consistent with previous linguistic investigations [20–22, 29]. ‘Sad’ was a word that was prevalent throughout, regardless of whether referring to feelings for oneself or for others, or feelings captured at the beginning or ending of a course about death and dying. Thus ‘sad’ feelings were universally linked to death. Nonetheless, the MOOC participants commonly chose words to express their perception of other’s feelings towards death that were considerably more emotionally negative (‘fear’, ‘scary’, ‘loss’) than the words chosen to express their own feelings towards death (‘inevitable’, ‘peace’, ‘natural’). When we moved beyond the self-report of words themselves and applied numeric sentiment scores to the words chosen, we found that the sentiment expressed to represent one’s own perspective was significantly more positive compared to those representing the perceived perspective of others. The words chosen to represent the feelings of others indicated sentiment that was more unhappy/unpleasant, more arousing/excitable, and more submissive/dominated by external forces. This aligns with findings from other studies [5, 59, 60]. Our findings leave us with the question of why the MOOC participants think that others feel so differently about death than they do themselves? Part of the explanation is likely to lay in the self-selected nature of our sample of people who chose to participate in a course about death, a considerable proportion of whom identified as health professionals and may have seen themselves as more informed and conditioned to death. But, the sentiment of the words chosen to represent the perceived feelings of others does demonstrate the assumptions people make about other peoples’ negative reactions regarding death. It is possible that these assumptions have implications for the willingness of people to start conversations about death and dying with the people around them. If we perceive that others will become distressed by bringing up the topic of death, are we less likely to attempt raising the topic? Does this avoidance then leave important things unsaid?
When we compared the types of words participants chose to describe their personal feelings about death at the beginning and the end of the MOOC, we found that the most frequently mentioned words were quite similar at both time points (e.g., inevitable, peaceful, peace, natural, sad), but that mentions of more negative words like ‘sad’ reduced considerably in frequency over time, along with a corresponding increase in mentions of words related to acceptance and comfort. The specific words chosen by individuals at each time point showed more changes over time than stability. When the sentiment of the words chosen to express personal feelings about death at both time-points was analysed, we found that participants chose happier valence words, less arousing words, and more dominant words at the end of the course about death than they did at the beginning. This finding suggests that participating in an online course about death and dying may have had a positive effect on the type of language people chose to express their feelings about death, with the emotional sentiment of the language used by the end of the course subsequently becoming more pleasant/positive, calmer, and more internally controlled. Given that the sentiment of words deals with emotion in relation to an issue, this suggests that over time the course may has assisted participants in becoming more emotionally accepting in their feelings and attitude towards death. It is also important to note therefore that spending time thinking and learning about death over a five-week period did not appear to have a negative emotional impact on course participants. This finding is consistent with our findings from research with our 2016 cohort using formal scale measures [6] and evaluation questionnaires [5] geared more towards behavioural indicators. The findings of the present study indicate that participation in the MOOC potentially had a positive emotional influence on the participants, in addition to the behavioural and cognitive influence demonstrated in earlier studies [5, 6]. This offers valuable triangulating evidence validating the potentially beneficial effects of participating in an online course about death.
Our multivariate findings demonstrated that the positive changes in sentiment occurred similarly for participants regardless of education, occupation, and location. However, an interesting influence of age was found on the rate of change in valence and dominance sentiment scores from the beginning to the end of the course. From the outset, younger participants scored lower on death sentiment valence and dominance, and compared to older participants, younger participants experienced greater change during course participation on the sentiment of the language chosen to express personal feelings about death, with greater increases in the pleasantness (valence) and dominance (power and control) in the words they chose at the end of the course. This may be in part due to younger participants having lower valence and dominance scores at baseline, meaning there was more opportunity for improvement in their scores by the end of the course. For older participants, the influence of a course discussing death is perhaps less due to having more personal exposure and proximity to death in their life and thus greater mortality salience. Therefore, it’s possible that participation in an online course about death may be especially beneficial for younger people, potentially assisting them to become less emotionally negative and more accepting of death and its inevitability.
Implications
The results from this study provide further validation of the emotive nature of death and dying. It provided quantification of the strength and direction of affective responses to death and the differential perceptions of death attitudes participants applied to others in the community. Word sentiment analysis was a useful adjunct to traditional self-reports, and is arguably less invasive.
The words and labels used in clinical encounters can have differing effects upon patients and clinicians. For example, Tayler and Ogden [81] reported the tendency of doctors to use euphemisms instead of the direct term ‘heart failure’ due to the dilemma of the latter causing more negative emotional reactions in patients. On the other hand, a study of more benign conditions (gastroenteritis/tonsillitis) found the use of clinical language over lay language can validate the patient’s sick role and can increase confidence in the doctor [82]. The use of euphemisms to avoid the word death is reportedly common, and accompanied by the potential for misunderstandings [20, 22, 58]. Thus, words are not neutral. They can have an impact on patients and people–they can bring people with us or alienate them, they can be interpreted differently by people, they can be delivered in personal or impersonal ways, and they can sway a person’s actions. There may be implications worth exploring for clinical settings, particularly for awareness of the choice of words verbalised in clinical encounters related to the end-of-life. Understanding the sentiment of words provides valuable insight into the emotional connotations tied to words we use, and could be useful for guiding clinical conversations in palliative care [20, 22].
The knowledge gained from this study about community death attitudes could potentially inform the creation of dialogue and messaging used in public health campaigns about the end-of-life preparedness. The use of higher valence and dominance words might bolster the persuasiveness of campaigns in a way that potentially leads to more effective emotional engagement and behavioural activation in target populations [37, 83].
Strengths, limitations, and future directions
Whilst this study was based on a large sample, several potential limitations require consideration. Like most MOOC samples, our sample is unlikely to be representative of the general population [62–64]. Participants were a self-selected sample of people in the community who chose to enrol in a course about death, and therefore may have been more likely to feel comfortable with the topic of death from the outset than people in the general population. Our sample also included a considerable proportion of people who identified as health professionals, which may impact on death attitudes given their greater opportunity for exposure and possibility of more desensitised death-related emotions. The sample was also comprised of a vast majority of females, which meant that gender comparisons could not be meaningfully undertaken. Furthermore, the results of the present study may be somewhat less representative of younger people and those residing outside Australia, because enrolees who subsequently went on to participate in the MOOC activities were slightly older and more likely to be located in Australia than those who enrolled but did not participate. Replication of the study with a representative community sample would be a valuable avenue for future research, to garner a stronger understanding of socio-demographic variations in the sentiment of words about death. Comparing death word sentiment for various age cohorts in the community with differing death exposure experiences is particularly needed, as well as direct dyad comparisons of self-other death attitude perceptions. It would also be informative to examine how respondents’ death word sentiment relates to their self-reported fear of death and other emotive death attitudes rated on standardised instruments. This could enable a broader insight into various aspects people associate with death [7].
A strength of this study was the inclusion of a longitudinal component that observed changes over time in the sentiment of words chosen to express feelings about death. However, course attrition meant that slightly less than half of participants that provided words at baseline also provided data at the end of the course. It is possible that those participants who were less inclined to complete the Time 2 activity were also less likely to have experienced a positive emotional response to the course. Nonetheless sensitivity analyses using alternative approaches (Multiple Imputation) to account for the missing data obtained similar results and conclusions to those using the complete-case sample. In the final week of the course when participants were asked to provide three death words again, it was possible for them to go back and look at their previous word choices from the beginning of the course. It is not clear how many participants went back and checked their first response, versus those who followed instructions to simply go with their instinctual response. It is not known what effect any comparative check by participants may have had on the words they chose to report at Time 2.
Of further importance is that our longitudinal assessment of change over time did not include a comparison group of people who were not exposed to the MOOC. Without this comparison group, it is impossible to rule out alternative explanations for the reasons word sentiment scores changed from the start to the end of the course. Other factors occurring within lives of participants while completing the MOOC may have caused changes on death sentiment over time that were not related to MOOC completion. Future research including a control group is warranted.
In this study, we applied sentiment analysis to three specifically chosen content words rather than to a written sentence, or a more generalised analysis of written text. However, it is possible that by giving participants explicit instructions we obtained more expressive emotional content words than we would have if a sentence was requested, given that typically only a small proportion of words used are classified as emotive [14]. The results may have differed if the activity used a sentence instead, but it may have reduced the emotive data elements provided, being more likely to have been limited to one core theme/lemma. Alternatively, utilising a much larger textual response may have provided a deeper understanding of emotions, as there is evidence that non-content function words such as prepositions and pronouns can also convey affective content [84, 85]. It would be worthwhile to explore the use of these different methodologies in future research, such as examining the effect of the online course over time by comparing the sentiment expressed in larger bodies of general comments/posts made at the beginning and at the end of the course.
Another potential limitation is that Warriner and colleagues’ [37] lemma database is based on rating words on an ordinal 1-to-9 scale, rather than a continuous interval scale. There may not be a uniform difference reflected between a rating of 1-to-2 as there is for 8-to-9. Very recently, new slide-rating methods for valence scores have been tested, which may provide a fruitful direction for future studies to utilise, if norms are developed [86].
The current study focussed on words used to describe feelings about death and dying, and has provided valuable insights into the way people feel about this issue. There is considerable potential for applying sentiment analysis to other related constructs, such as words chosen to describe feelings about Palliative Care, Advance Care Planning and Voluntary Assisted Dying. Such future research may help us better understand the nuances in community attitudes towards these important social policy issues. The possibilities of sentiment analysis as a methodology to detect emotional states of respondents also deserves future research attention. Previous studies have successfully used sentiment analysis to detect mental health conditions and level of threat in social media statements [17, 32]. Future research could pursue word sentiment as a method to detect emotional distress in health settings (e.g., carer bereavement risk), which could activate early intervention efforts.
No comments:
Post a Comment