Mental experiment in which a person was split into two continuers: Most took decisions based on the continuity of memory, personality, and psychology, with some consideration given to the body and social relations

Putting your money where your self is: Connecting dimensions of closeness and theories of personal identity. Jan K. Woike, Philip Collard, Bruce Hood. PLOS, February 12, 2020. https://doi.org/10.1371/journal.pone.0228271

Abstract: Studying personal identity, the continuity and sameness of persons across lifetimes, is notoriously difficult and competing conceptualizations exist within philosophy and psychology. Personal reidentification, linking persons between points in time is a fundamental step in allocating merit and blame and assigning rights and privileges. Based on Nozick’s (1981) closest continuer theory we develop a theoretical framework that explicitly invites a meaningful empirical approach and offers a constructive, integrative solution to current disputes about appropriate experiments. Following Nozick, reidentification involves judging continuers on a metric of continuity and choosing the continuer with the highest acceptable value on this metric. We explore both the metric and its implications for personal identity. Since James (1890), academic theories have variously attributed personal identity to the continuity of memories, psychology, bodies, social networks, and possessions. In our experiments, we measure how participants (N = 1, 525) weighted the relative contributions of these five dimensions in hypothetical fission accidents, in which a person was split into two continuers. Participants allocated compensation money (Study 1) or adjudicated inheritance claims (Study 2) and reidentified the original person. Most decided based on the continuity of memory, personality, and psychology, with some consideration given to the body and social relations. Importantly, many participants identified the original with both continuers simultaneously, violating the transitivity of identity relations. We discuss the findings and their relevance for philosophy and psychology and place our approach within the current theoretical and empirical landscape.

Study materials, supporting analyses, and qualitative coding scheme: https://doi.org/10.1371/journal.pone.0228271.s001

Excerpts of discussion...

The fission scenario as scientific fiction: Benefits and validity concerns

Not all philosophers embrace the use of hypothetical scenarios in studying identity [129–131]. Some are critical of fission thought experiments, arguing that the intuitions derived from such fantasy accounts—in which anything goes—violate natural worlds and are therefore not valid measures of how people conceptualize identity in natural settings (e.g., [132]). Like contaminated test tubes in biochemical experiments, results obtained with faulty scenarios would have to be considered dubious [131]. Scholl [130] further argued that reactions to “bizarre scenarios” with forced responses might “tell us more about heuristic getting-through-the-experiment strategies than about actual metaphysical intuitions” (p. 580). One could compare these scenarios to visual illusions that generate experiences that are at odds with reality—but these experiences nevertheless often provide insight into the mechanisms that generate the illusion. For example, size and distance illusions reveal the computation the brain uses to calculate physical dimensions even in normal cases. Likewise, intuitions derived from cognitive processes during these unreal and outlandish examples may not be necessarily meaningful when measured against the constraints of reality [130] but rather cast light on how we use and reason with the concept of identity. While the participants’ open answers showed a degree of confusion in a minority of subjects, most answers reflected a well-considered and principled approach to the questions. Studies in experimental philosophy often feature unusual scenarios, and replicability in this sub-field compares favorably with replication rates in other domains of psychology [133].

Our science fiction scenario might be criticized for its lack of realism. It makes crucial assumptions that contradict the scientific understanding of how the components that we neatly separated in the story interact and co-depend on each other. According to Harle [68], a brain cannot be considered to be independent of the body; it would immediately adapt to a new body, rework the inner representation, and react to changes in social responses. Brain and body are interdependent in complex ways [39, 51, 63]. Motor learning [134] also involves two components, body and memory; muscles cannot work without neuronal input. The body and appearance component used in our studies could be understood to either include or not include brain matter. If included, all other components would depend on this component, as there can be no psychology without consciousness. If, alternatively, higher mental functions and consciousness are regarded as part of psychology or memory (which seems to be the approach taken by at least some participants), the brain, as a bodily organ that makes psychological functioning possible, can still be seen as source of tension in the scenario framework.

Participants responded to our scenario only from a first-person perspective. First-person evaluations of identity and survival might differ from third-person evaluations [128, 135]. For one thing, many legal, practical, and social concerns can be fulfilled by a person who is a spontaneous true copy of the original person. How much participants care about the disruption of continuity might therefore differ depending on whether the replaced person is them or a neutral other person [30]. Rorty [18] distinguished between an external observer’s perspective on individual identification and an individual’s internal perspective; features essential to an individual’s self-perspective might be irrelevant for an observer, and not all philosophers assume that first-person judgments have final authority [136, 137]. On the other hand, at least one study explicitly testing for the effect of perspective on intuitions about identity based on [28] found no substantial differences between a first-person and a third-person perspective [31]. At the same time, continuers were introduced from a third-person perspective. This shift was necessary to avoid a pre-judgment of the question how the original person relates to them, but might create a perceived distance. This could make it easier to give a negative answer to the survival question, but would have a symmetric influence on identity responses for both continuers.

It is less clear which perspective is better suited to evaluate claims about identity and survival. We assumed that involving the participant would be the best way to increase attention to the situation and diligence in responding. We induced a connection between our categories and their real-world instances, as perceived by the participants. A participant evaluating the importance of “body” in the money allocation problem will thus take the perception of her own body into consideration, which may lead to different results than when considering the importance of a body. This fact might play into our finding that a subgroup of participants placed a negative value on the continuity of the body.

On the other hand, our scenario avoids a complication common to many other fission scenarios: By splitting possessions and friends between survivors, it sidesteps the problem of nonsharable singular goods [34] with symmetric claims from two sides. These claims include access to special objects and, more significantly, the chance to engage in special relationships with others [51]. On the flipside, as Schechtman [138] argued, the scope and time frame of fission scenarios does not allow for societal reactions to the products of fission, which might include changes to the concept of persons and identity. Our approach avoids the confound of money being allocated for reasons of loss or pity, as both survivors emerge with their lives, bodies, and environments fully intact (if not unchanged). Williams [28] predicted a framing effect in the understanding of the scenario, a type of “leading the witness,” with a reversal of intuitions regarding identity depending on whether the story is told as involving body-swapping (transferring a person’s memory to a new body) or mind-manipulation (creating new memories in a person’s body). Empirical work has confirmed these predictions [31]. To counteract such potential framing effects, we used parallel language for all components varied in continuers; none of the components was privileged in the description of the scenario. Further, similar tensions involved in our scenario are discussed in section 1.3. in S1 Supporting Information.

Invoking hyperspace travel buys some degree of freedom, but at the cost of physical implausibility. However, factual impossibility does not prevent imaginability, and as Johnston [139] argued, “such per impossible thought experiments might nonetheless teach us about the relative importance of things that invariably go together” (p. 601). It is precisely because some aspects are not easily separated in realistic scenarios that we chose a fantasy scenario, allowing us to explore intuitions whose tests would otherwise be confounded.

Our scenario remains within the conventions of popular films (e.g., Total Recall or Blade Runner—both of which fittingly now exist in two versions) that deal with cases of copied or artificial memories and identities [140]. Body and mind transfer were considered intelligible by Locke [9], and the nature of personal identity is a recurrent theme in literature. Many readers have appreciated Franz Kafka’s tale of Gregor Samsa’s sudden metamorphosis into an insect [12] or Ovid’s metamorphosed subjects who survive the transformation [39]. Children become acquainted with bodily transfer in fairy tales like The Frog Prince [23, 120] or Hans Christian Andersen’s The Little Mermaid [39]. Johnson [141] takes this mere imaginability as an argument against declaring bodily continuity as a logical precondition for personal identity. Our scenario is no more fantastic than other thought experiments that have been employed to disentangle identity from its natural correlates—through neurosurgeons [6, 38, 51, 128, 142], amoeba-like duplication [127, 143], cloning [144], parallel universes [30], or even swamp-beings [66]. To come full circle, some of the philosophical conceptions and puzzle cases are reproduced in cultural creations and thereby further embedded into cultural consciousness [39]. It is possible that there is no theory of personal identity that would be able to “satisfy all intuitions about all devisable scenarios” (in [31], p. 297), but the advantage of imaginary scenarios lies in their power to isolate phenomena, which makes it possible to attend to specific aspects of our concepts [145]. In our scenarios we can separate changes from their ordinary causes and study decisions that may not occur in the world, but that probe concepts we apply to the world. In addition, our use of a novel paradigm limits the danger of previous exposure to similar questions and potential confusion, which is a concern with crowdsourced participants [146]. We therefore maintain that there is value in our approach of pitting dimensions that are generally accepted as dimensions of closeness against one another.


Dimensions of closeness

Our scenario bundles features into dimensions that might be further differentiated. For example, in contrast to [65], we did not differentiate between personality and psychology, on the one hand, and moral values, on the other. Evidence from several studies considering real-world personal transformations has indicated that identity judgments are most heavily influenced by changes or non-changes in moral values [65]. Changes in morality were judged to be more relevant than changes in (non-moral) personality attributes or memory. In a similar vein, Strohminger and Nichols [116] found that changes in morality in patients with neurodegenerative diseases strongly determined changes in perceived identity. Nunner-Winkler [21] reported on a study asking participants which changes would lead them to see themselves as a different person. Ideas about right and wrong and sex membership were considered to be quite important; appearance and money were considered less relevant (although some participants rated looks to be important, consistent with our distributional results).

The distinction between moral and nonmoral traits is somewhat ambiguous (e.g., conscientiousness was considered as a moral trait rather than a personality factor in [65]). One person’s morals do not and cannot exist in a social vacuum, moral consensus is central for co-ordination, affiliation and conflict resolution. Morality stands in complex relations to beliefs, values, behaviors and communities. It also depends on memory in nontrivial ways. Some of the induced changes in the scenarios even involved the loss of the moral faculty with a likely ripple effect reaching other dimensions of the self. If this perspective is true, the relevance of morality for personal identity might lie in these possibly disruptive consequences of changing one’s morals in relation to one’s environment and not because of its self-defining importance. Evidence for this interpretation is found in two studies demonstrating that changes in widely shared (and therefore less unique to the individual) moral values are considered to lead to more changes to the person than changes in controversial moral beliefs [106, 147]. For controversial moral beliefs, that might be considered most defining and informative for describing a person’s self, the effect was weaker than for memory. Also, the changes in memory induced by our scenarios would induce both errors of omission and errors of commission, which can have differential impacts on moral behavior [148]. In contrast, some studies focus mostly on omission errors due to memory changes (e.g., [118]), which are describes as having more limited effects on behavior towards others than changes in morality. In our scenarios, dimensions are replaced by random sampling from the participant’s reference population, which is a different operationalization of change. Heiphetz and colleagues [147] showed, for example, how the perceived change was mediated by perceived disruptions of friendships.

Some argue that morality is not even conceivable without personal identity [6, 10, 58]. Most people also seem to have inflated beliefs of their own morality [149]. In separate evaluations, our participants ranked memories and psychology to be more important for identity than moral values. Nonetheless, a further decomposition of the broad headings we used in our study would be feasible and interesting in future research. In particular, the role of moral traits and behavioral tendencies could be considered separately, even within a similar factorial setup as the one we employed.

The social dimension could be further differentiated, as well. Parents were considered to be more important than friends in Study 2, and Nunner-Winkler [21] reported similar findings. Of course, parents influence a person directly through the transmission of genes as well as indirectly through instruction and parenting behavior; changing one’s parents cannot be considered a merely social manipulation and could well have an impact on every other dimension.

In our scenario, changes in memory are considered universal and all-or-nothing. In real life, however, memories of self or self-knowledge seem to be better preserved than other knowledge, even in semantic dementia [20], and a subjective belief of self-persistence is demonstrated by patients with Alzheimer’s disease [150]. Alternatively, the sense of self may be impaired while episodic memories stay intact—as in the case of R.B. [67]. Further, separating specific psychological aspects or memory from a person’s social context and network of activities might prove impossible in practice [18]. There is also some overlap between criteria based on psychology and memory, but under the assumption that two organisms with the same memories might nonetheless differ in personality and psychology (e.g., based on differences in needs, intentions, values, or goals), it is not necessary that the criteria coincide. In fact, the psychological continuity criterion has been proposed as a critique of a narrow Lockean focus on memory [11].

Critics of our scenario might further object that our random collage of features in the two continuers destroys the causal connection between past and present states necessary for identity [6, 30]. Preschool children already individuate objects and persons spatio-temporally [23, 151] and, following Sagi and Rips [152], causal histories receive special attention in linguistic disambiguation in discourse. In all our scenarios (except the two extreme cases with exact duplicates), change in characteristics was induced by an accident, an unusual life event that disrupts spatio-temporal continuity. This fact might strengthen impressions that identity is not preserved. According to data reported in [21], for example, participants regarded changes in attitudes or beliefs that were due to normal life experiences as non-consequential for identity judgments—as opposed to changes induced by brainwashing, severe medical conditions, or accidents. Therefore, the nature of the transformation might play a role in our participants’ judgments. Note that both continuers underwent the same procedure, so this factor cannot explain differential assessments. Although the abruptness and symmetry of the original person’s transformation prevents the application of spatio-temporal continuation criteria, participants might still construct “fictive causal histories” [153] to assess which of the two continuers might have the better chance of being the result of changes within an ordinary life.

Finally, for a continuer to acquire a random set of possessions, these would have to materialize from somewhere. Our scenario also assumes that this change in possessions can leave memory, psychology, friends, and appearance untouched. This is incompatible with the reality that some of our memories are intertwined with objects in our possession and the difference between owning or not owning status symbols, for example, can impact self-value, build and burn bridges with others, and change perceptions of their owner.


Towards a process model of re-identification

Our studies allow to make some progress in the analysis of decisions involved in determining personal identity. Like Rips and colleagues [17], who develop the causal continuer model based on Nozick’s theory, we are interested in the decision process. Decision processes, as implemented by human beings, are often insufficiently described by functions merely predicting decision outcomes. A further analysis of the decision processes needs to address questions of information search: Which persons are considered as continuers? When and why is the search for possible continuers stopped? Which dimensions are considered in the subjective closeness metric, and how are these dimensions integrated? We showed some results compatible with decision-making following the closest continuer logic. Is there further evidence for the three steps being followed in a specific sequence—the fast-and-frugal tree in Fig 1 would not yield different outcomes if the first three levels changed their relative position—and how stable is this process across individuals? A structurally similar model of decision making has been proposed for explaining the phenomenon of choice deferral [154, 155]. When faced with a selection of possible alternatives, choice in the 2S2T-model [155] is deferred for one of two reasons. First, none of the options is good enough and surpasses a decision threshold or second, too many options are good enough, surpassing the threshold but it becomes difficult to choose the best option. Of course, personal re-identification is not simply preferential choice but the analysis of the decision process might still be informed by the analysis of related or parallel processes in other domains.

While the mathematical form of weighted-additive linear models implies weighting and adding, many other operations, such as lexicographic stepwise procedures that ignore (sometimes most of the) variables in the equation [35] would still be captured by this model [156]. Brook [12] argued for a model of personal re-identification that starts with psychological factors and only considered other dimensions if the information is missing (or inconclusive). Variance in choosing and applying criteria might again be related to other individual differences [41, 42]. Based on the variations in our chosen design for this study, it is not yet possible to build cognitive models of participants’ decisions. It is, for example, unclear whether an appropriate model should be stochastic, as in [17], or deterministic.

Our scenarios varied factors that should mostly influence the assessment of closeness and only indirectly the decision-making based on these assessments. Future research could shift this focus to the subsequent stages of the procedure. Thus, specific exit nodes of the decision tree in Fig 1 could be investigated. For example, is there a minimum level of closeness required for participants to determine that any of the continuers is identical to the original person? Do participants share the intuition that a fission resulting in multiple exact copies does not preserve identity, and would this depend on the level of closeness? What type of difference is considered to be sufficient to single out a closest continuer?

Both studies in this manuscript confronted participants with two continuers. Future studies could increase the number of continuers. A different approach could focus on one continuer by either keeping a second continuer constant, or moving from paired comparisons to binary reidentification. Previous studies have explored variants of thought experiments compatible with these ideas. White [157] implemented one such scenario, focusing on the likelihood that a living person might be the reincarnation of a deceased person (see also [65]). For reincarnation judgments, distinctiveness was found to guide decisions. Similar to our sci-fi scenario, this setting might introduce specific assumptions about the process of reincarnation that could guide responses. For example, the importance of body similarity might be evaluated to be a lot lower than when responding to a scenario, in which a ship wreck survivor is returned from an island and matched to missing persons, and the importance of moral attributes to be higher. In contrast to the second scenario, the reincarnation scenario prevents the use of causal histories that are useful for person tracking [1].

It might also be the case that different practical concerns demand different criteria of identity. We investigated the parameters of identity in the context of re-identification and compensation. Other practical concerns, such as attributing blame, responsibility, or guilt, or allocating punishments and rewards, might trigger different responses, as the criteria of identity might shift or current properties of persons might become more relevant than historical properties and re-identification questions [10, 158].

To sum up, while we made progress to shed light on the decision processes used by participants, we have not yet established a complete process model, which should be the goal for future research [159, 160].


Does it matter what matters for reidentification?

What are the implications of our studies for debates in cognitive sciences and other disciplines? Lay intuitions may be more prone to error than those of philosophers [100]—although experts’ authority may also be questionable, as their philosophical intuitions are partly a function of their personality [161]. Our participants’ endorsement of identity relations of an original with two non-identical persons violates the transitivity of identity. Yet this response pattern may simply show that our participants do not conceive of personal identity as strictly numerical, or that they have alternative conceptions of persons. Our results are in any case highly relevant for a descriptive analysis of people’s understanding of identity and their theories of survival. Further, Nozick [14] would not attribute the variance in people’s perspectives to errors or false everyday beliefs [162], but rather to the variance in closeness metrics legitimately deemed appropriate by different persons. Our findings are similarly relevant for cases in which scientists, philosophers, or marketers try to appeal directly to lay intuitions or common sense.

Philosophers and psychologists differ in their conceptualizations of intuitions [163], yet in our complex scenarios participants could not arrive at their answers without careful assessment. When appealing to the common sense of people both in theorizing and in legitimizing operationalizations, a researcher “should respect what ordinary people in fact say when asked—unless they are somehow led astray” (in [115], p. 216). Study 2 eliminated one way in which participants may have been led astray, by moving from a continuous to an all-or-nothing decision. Any appeal to general intuition should take into account both our main results and the interindividual variability demonstrated in both studies. Any attempt to measure the conceptualization of self or personal identity may be informed by both our positive and our negative findings.

Psychological research has analyzed psychopathological conditions that entail potential breaks in personal continuity and identity [164]. For example, patients with the Fregoli delusion are convinced that different people are in fact a single person appearing in a variety of disguises. Here, the recognition of the outward appearance is separated from the identification of the person. This disorder goes beyond prosopagnosia, or face blindness, where the perception of faces does not allow for identification of persons (see also [165]). Patients with Capgras delusion believe that a specific person, often a loved one, has been replaced by a duplicate who is indistinguishable from the original person. In cases of mirror misidentification, patients fail to recognize their own reflection and infer that the person in the mirror must be someone else [166]. These observations from clinical psychology indicate that the neural mechanisms of identification cannot be reduced to acts of perceptual recognition, and hint at the requisite capacities necessary for personal reidentification; furthermore, understanding ordinary reidentification processes might help to understand and locate their disruption.

As the introductory example shows, the importance and reach of identity questions are not limited to specialized academic discourse, even if not every instance is as dramatic as the hanging of Arnaud du Tilh. A survey of current debates outside philosophy referencing the personal identity literature creates the impression that many of Parfit’s [6] suggestions, examples and ideas are still in the process of being (re)discovered. Mott [167] explored Parfit’s suggestion that diminished personal connectedness might be a reason for statutes of limitations (p. 325), and provided evidence that desert of punishment and grounds for criticizing a person for past deeds are considered to diminish over time, which is partially explained by the reduction in closeness. A second debate is centered on the question of the validity of living wills after substantive changes to a person’s cognitive capabilities. For example, can a competent person impose values and interests on the future incompetent person or should the strength of advance directives grow weaker with the loss of closeness [21, 81, 168]? The incompetent person might, for example, derive unexpected pleasure and satisfaction under conditions the competent person did not foresee. Further, references to a future self might have a tremendous impact on ethical behavior [46], motivation, and goal-pursuit [169]; the sense of temporal persistence can motivate future-oriented self-regard and short-term sacrifices benefiting future outcomes [43, 44]. Without personal identity, it would be meaningless to make promises, grant ownership or the right to vote [79], or offer compensation [10, 12]; challenges to personal identity affect the institutions built upon it.

Disruptions of personal identity have been shown to severely impact people’s lives. Chandler and colleagues [61] presented impressive evidence of the connection between the inability to give an account of one’s identity and the risk of adolescent suicide, and how cultural continuity can moderate the elevation of suicide risk in vulnerable minority groups. This line of research raises the question or how one’s perspective on personal identity is shaped by the social environment and connects the concept to mental health. A diminished sense of self and the self’s stable existence is also deeply intertwined with borderline personality disorder [170]. Furthermore, challenges to personal identity can also emerge due to technological innovation. Notions of identity are fundamental in conceptualizing behavior in virtual environments [171] and have implications in law in connection with identity theft [172] or impersonation. Pascalev and colleagues [63] discussed how first suggestions for a medical head transplant procedure introduced questions of personal identity into neuroethics.

The gravity of these real-world examples goes well beyond that found in hypothetical thought experiments. The analysis of contrafactual scenarios has nevertheless paved the way for addressing real-world concerns and situations, whose connection to personal identity is discovered through analogy, created through technology, or bestowed by social institutions. Understanding how the concept is perceived and applied and how experimental puzzles of seemingly little direct relevance are tackled and solved can ultimately inform practitioners and theoreticians facing recurrent and novel situations with serious consequences.