1 Introduction

Typically, risk is the likelihood of an undesirable event and involves two main components: the probability of something going wrong and the negative consequences if it does. The assessment of risk is crucial in decision-making processes in everyday life and for different professions. Clinical psychologists assess a patient’s suicide risk by evaluating the risk level. This involves identifying specific risk factors such as mental illness, previous suicide attempts, family history of suicide, substance abuse, and stressful life events (van Heeringen, 2018). The aim is to determine the immediacy and severity of the threat to appropriately intervene and provide the necessary support and treatment.

Recently, however, there is doubt about the usefulness of suicide risk assessment. Meta-studies show that suicidal action is currently unpredictable, and psychologists cannot use risk factors to reliably distinguish between high and low risk patients (Franklin et al., 2017). A meta-study conducted by Matthew Large and his team (2016) shows that current approaches cannot differentiate between high-risk and low-risk groups. The study reveals that roughly half of all suicides occur in lower-risk groups, and more than 90% of high-risk individuals do not attempt suicide. Practically, the high false positive rate renders suicide risk assessment as useless (Nielssen et al., 2017).Footnote 1

In this manuscript, I develop an outline of a new understanding of suicide risk that does not depend on the probability of an event and, therefore, on predictive abilities. In contrast, according to the Normic account of suicide risk, the more normal an event is, the higher the risk. However, normalcy should not be interpreted as statistical regularity, because then theorists would return to the probability of an event. Normalcy is contingent upon certain norms. I argue that norms of practical rationality can render suicidal action more or less normal, depending on an individual’s reasons for living and reasons for dying. For example, if someone who has no reasons for dying and who has reasons for living attempts suicide, this act is abnormal. The divergence from normalcy can be discerned by theorists through the amount of explanatory effort required to make the action intelligible. To comprehend an unreasonable suicide attempt, theorists must depend on complex causal theories.

Given limited resources for helping people, psychologists prioritize high-risk patients. Traditionally, the goal was to prevent suicide in those likely to act on suicidal thoughts. However, due to the difficulty in accurately predicting suicidal behavior, the Normic account of suicide risk focusses on those who perceive the most reasons to take their own lives. These individuals are identified as needing the most immediate help.

In Sect. 2, I will present data suggesting that suicidal action is currently unpredictable. In Sect. 3, I explain why this unpredictability is problematic for the conventional understanding of suicide risk. Traditionally, individuals with a high risk of suicide are those most likely to exhibit suicidal behavior. In Sect. 4, I introduce the Normic account of risk. Finally, in Sect. 5, I develop a Normic approach to suicide risk, which determines the normalcy of an action by referencing norms of practical rationality. In Sect. 6, I briefly discuss theoretical implications for understanding suicide risk with the Normic approach.

2 Low predictive powers and the crises of suicide risk

A meta-analysis conducted by Joseph Franklin and his team (2017) revealed that despite 50 years of empirical research on suicide, the ability to predict suicide did not improve. This meta-analysis examined empirical findings from 1965 to 2014. Surprisingly, all recognized psychological risk factors demonstrate only slightly better predictive abilities than random chance. Forecasting future suicidal thoughts and behaviors through a random coin toss would yield predictions with approximately 50% accuracy. The meta-analysis highlights that none of the known risk factors can predict significantly better than with 60% accuracy. Chan and colleagues (2016) warn to rely on known risk factors, as they potentially mislead mental health professionals. Belsher and colleagues (2019), taking machine learning studies and access to mental health databases into account, conclude that the “accuracy of predicting a future event is near 0” when suicide prediction models are considered.

In a prospective study, Alex Pokorny examined 4800 men and rated 803 (17%) as high risk of suicide. Over the next 5 years, 30 (3.7%) high-risk patients died by suicide, leading to a proportion of false positives of 96.3%. In another prospective study from Goldstein (et al., 1991) with 1906 patient, not a single of the 46 suicides occurred among the high-risk group. More recently, a prospective study from Madsen (et al., 2012) revealed a suicide rate of 0.14% among the high-risk patients, while Steeg (et al., 2012) reported a suicide rate of 0.5% in their prospective study for high-risk patients. Alex Pokorny concluded that the high false positive rate renders suicide risk assessment as practical useless (also, see Nielssen et al., 2017).

Based on the data, suicide risk assessment is ineffective, as there is no reliable way to predict suicidal behavior. Although clinical psychologists routinely endeavor to identify individuals at risk of suicidal behavior, all known risk factors prove insufficient for this task. This does not imply that mental health professionals should cease using these risk factors. Statistically, there exists a subtle difference between groups of individuals with and without these risk factors. However, when it comes to identifying specific individuals at risk, these risk factors offer no utility. A meta-study conducted by Matthew Large and his team (2016) shows that current approaches cannot differentiate between high-risk and low-risk groups. The study reveals that more than 90% of high-risk individuals do not attempt suicide. Importantly, half of all suicides occur in lower-risk groups showing the risk classification is misleading. Also, that there is “no evidence that the statistical strength of suicide risk assessment has improved over time” (p. 12). Regrettably, intuitive predictions made by professional clinicians regarding whether an individual will attempt suicide are also unreliable (Nock et al., 2010; Swets et al., 2010).Footnote 2

If suicidal action is currently unpredictable, what does it mean to have a high or low suicide risk? All referenced studies assume that suicide risk is synonymous with the probability of suicidal behavior. Therefore, if the probability of suicidal behavior is uncertain, the concept of suicide risk becomes vague.Footnote 3

3 What is ‘risk’?

There are, at least, two ways to talk about risk, the everyday concept and the technical concept of risk (Hansson, 2023). On one hand, we employ the everyday concept of risk in a vague yet effective manner. For example, we might caution someone, “What you’re doing is risky,” or “There’s a high risk of you slipping on this wet hiking trail.” In such instances, we apply the term ‘risk’ to situations where an undesirable event is possible but not certain. This is often referred to as subjective risk. On the other hand, there exists a technical concept of risk. The technical usage often pertains to the probability of an unwanted event, sometimes termed as objective risk. For example, a health professional might explain to a client that smoking increases the risk of lung cancer. In this scenario, lung cancer, the undesired event, has a certain probability ‘X’ of developing in an average individual, but this probability rises to ‘X + n’ if the individual smokes. However, here people usually discuss the statistical expected value of an undesired event occurring. The expected value of a negative event is the product of its probability and its negative impact.

Pritchard (2015) has argued that the technical term of risk, which refers to probabilities, is too limited to account for all relevant cases. In fact, Pritchard is not the only philosopher who considers the technical term of risk as too narrow (see Buchak, 2014; Shrader-Frechette, 1991). Pritchard challenges the probability-based view of risk using a thought experiment. Imagine there’s a bomb that will explode if a specific combination of numbers is drawn in the national lottery– the odds of this happening are fourteen million to one (bomb 1). Now, imagine a similar scenario with the same odds, but this time the chance doesn’t come from a lottery draw but from a highly improbable series of events, like the weakest horse winning a race by at least ten furlongs (bomb 2). Pritchard posits that, in a certain sense, the risk associated with each event is different, even though the probabilities of the bombs exploding are identical in both scenarios. In fact, an empirical study supported Pritchard’s hypothesis that people would prefer bomb 2 over bomb 1, despite the equal probabilities of explosion (Ebert et al., 2020). This suggests that our perception of risk is influenced by factors beyond mere probability, potentially involving the nature or context of the risky scenario.

One way to look at the risk-conception, is to assume that the technical view of risk is the only right one, namely the objective risk. From this monist conception of risk, there is only one real risk, which grounds on the probability of an unwanted event. Amos Tversky and Daniel Kahneman (1974) famously considered systematic divergences from the probability view as a bias– a failure in human reasoning. For instance, is there a higher risk for being killed by a shark or by falling airplane parts? Most people’s first reaction is to consider the risk for being killed by a shark higher. In fact, it is 30 times more likely to die by falling airplane parts (Death Odds, 1990, September 24). Tversky and Kahneman (1973) explain such deviations from the real probability with the availability heuristic. A heuristic is a mental short-cut to solve a problem, which is, in contrast to an algorithm, not always up to the task (Gigerenzer, 2007). The availability heuristic in particular relies on immediate examples that come to a person’s mind. Because shark-attacks are something that is represented in the media, it is more likely to come up in people’s mind. Therefore, so their reasoning, it must be more likely to die by a shark.

Philip Ebert, Martin Smith and Ian Durbach (2020) argue for a pluralist view of risk: “there are multiple, equally legitimate, notions of risk.” (p. 447) They suggest that different understandings of risk, including the probability view, stem from the intuitive folk notion of risk, while emphasizing different aspects. From the pluralist perspective, divergence from the probability calculation of risk does not automatically imply reasoning that leads individuals astray. Although biases and failures in risk assessment are still possible from the pluralist viewpoint, a divergence from the correct probabilities is not sufficient to indicate a reasoning error. Ebert, Smith, and Durbach themselves develop the Normic account of risk, which I will introduce in the following.

4 The normic account of risk

Similar to the modal account of risk from Pritchard (2015), the Normic account of risk from Philip Ebert, Martin Smith and Ian Durbach (2020) makes use of the possible worlds approach instead of relying on probability calculations.Footnote 4 The determination of closeness between one world and another is based on how normal or abnormal it would be for a fact to occur. Therefore, closeness between possible worlds is not defined by their similarity. From this viewpoint, Pritchard’s bomb 1 appears riskier than bomb 2: It is more normal for a set of numbers to be drawn in the national lottery than for a series of abnormal events to occur. The latter requires an explanation to make the occurrence of such abnormal events understandable, which is not necessary for the trigger of bomb 1, where six specific numbers are drawn in the next national lottery. This happens because it is a normal event, needing no explanation to make sense of its occurrence.

What, then, constitutes ‘normal’? Responding to this question by referencing what is statistically normal is an unsatisfactory solution. If the frequency of an event is the determinant of its normality, then we are essentially reverting to the probability view of risk. Merely using a different term, ‘normal’, to describe a high frequency of an event’s occurrence does not introduce a significant distinction.

The Normic account does require a definition of what is considered normal. To illustrate, consider the case of determining the risk of dying from a shark attack versus falling airplane parts. According to the Normic account of risk, the risk hinges on which event is more normal. But if this assessment is not based on the frequency of such events, what else can we refer to?

Generally, the normalcy of an event can be influenced by various factors. One factor is natural laws. Even if an event is highly improbable, if it results from a natural law, it can be deemed normal in a certain sense. In medicine, ‘normal’ often refers to a range of values or conditions considered indicative of health or proper functioning. For example, when a doctor says a heartbeat is normal, it implies that the heart’s rhythm supports its proper functioning, indicating no abnormalities. Clinical psychologists also utilize this concept of normalcy, when, for instance, they view certain aspects of grieving as normal, while others are deemed abnormal in the sense of indicating dysfunction.

These sketches of what can determine the normalcy of an event raise many questions. However, as it will turn out, the aim of this paper is not to offer a comprehensive account of normalcy but to focus on explaining the normalcy of suicidal actions with the norms of practical rationality.

5 Normic suicide risk: a new suggestion

In the previous section, I introduced the Normic model of risk, which is based on a possible world framework. In this framework, various ‘worlds’ are conceptualized, each representing a different way things could potentially unfold. The Normic model assesses risk by evaluating how a given event fits into these possible worlds, with a particular focus on the ‘normality’ of the event in these scenarios. The closeness between possible worlds is determined by how normal events are: “According to the Normic account of risk, the risk of a proposition P is determined by the normalcy of the most normal worlds in which P is true–the more normal these worlds, the higher the risk” (Ebert et al., 2020, p. 444).

The risk associated with an event increases with the event’s normalcy. It is important to remember that the level of risk depends on the explanatory effort needed to make the occurrence of an event intelligible. Accordingly, if a theorist adopts a measure of normalcy, given its normative dimensions, the risk level for an event’s occurrence depends on how much explanatory effort is needed to explain that occurrence. If little explanatory effort is required to explain an event, then the risk for that event is high. In contrast, if a substantial amount of explanatory effort is needed to make sense of an event’s occurrence, the risk is low.

5.1 Explaining intentional action

Suicide is intentional self-killing (Cholbi, 2011), and there are different ways how to explain intentional actions. Broadly speaking, theorists can approach action explanation from either a psychologistic or non-psychologistic standpoint. In the psychologistic approach, an agent’s mental states, such as psychological attitudes and emotions, serve as explanatory factors for their actions. In contrast, the non-psychologistic approach relies on reasons (or goals) to explain actions. Regardless of the chosen approach, theorists seek answers to Why-questions (Anscombe, 1957), such as ‘Why did John go to the hospital?’ No matter which approach for action explanation is favored, the relevant norms are the norms of practical rationality. The explanation reveals the light in which an agent viewed the action as choice worthy. The explanation “[…] leads us to see something the agent saw, or thought he saw, in his action– some feature, consequence, or aspect of the action the agent wanted, desired, prized, held dear, thought dutiful, beneficial, obligatory, or agreeable” (Davidson, 1963, p. 685). Regardless of whether a theorist prefers the psychologistic or a non-psychologistic approach, explanatory theories for intentional actions rely on the norms of practical rationality. Once an action is explained, the action makes sense as it became intelligible (Paul, 2020).

From a psychologistic view, theorists can explain actions by pointing to the agent’s psychological features that reveal the agent’s understanding of the action and the underlying motivations. For instance, John went to the hospital, because he believes that his father is in the hospital and he wants to visit him. The idea behind such a psychologistic action explanation is that actions are explained by citing the psychological features that caused the actions to occur. Donald Davidson, the most prominent advocate of this approach, suggested that an agent’s pro-attitude, consisting of a belief-desire pair, explains actions, as they are the underlying causes (Davidson, 1980). Causal explanations of action can rely on more complex mental states than beliefs and desires, such as mental maps, frames, or metal files. What is important is that the explanatory power of psychologistic explanations stems from causality. It is the fact that the mental states in question cause an action that generates the explanatory power. Therefore, philosophers of action call this explanatory approach the Causal Theory of Action.

However, the Causal Theory of Action suffers from two problems. Firstly, the Causal Theory of Action presupposes that the causing mental states are known. As Sect. 2 illustrated, psychologists cannot predict suicidal action, which raises doubts about the idea that psychologists know the causes of suicidal action (Bryan, 2021; van Heeringen, 2018). Psychologist know about risk factors, which are assumed to increase the probability for suicidal behaviors. However, studies show that these factors cannot predict suicidal behavior. Accordingly, if the mental states that cause suicidal action are unknown, an action explanation that relies on the causing mental states seems inappropriate.

Secondly, deviant causal chains pose a problem for the Causal Theory of Action (Asma, 2023). In certain situations, a bodily movement can be caused by a belief-desire pair but still not qualify as the agent’s action. Consider, for instance, a climber hanging on a wall who believes she has excessive weight in her pack and desires to drop something from it. While she possesses the appropriate belief-desire combination, this pairing causes her to become nervous. Due to her nervousness and body-shaking, she accidentally loses some of her mountaineering equipment. Deviant causal chains show that an action explanation depends on more than the causation of mental states representing one’s reasons for acting. What is needed, in addition, is the right way of causation. Intuitively, the right way of causation is ‘whatever it takes to act for a reason’, which raises doubts about the relevance of causation (Paul, 2020). It is a debatable prerequisite that ‘acting for a reason’ is a causal relation between an agent and the reason. If acting for a reason is the right kind of ‘causation’, then there seems to be something non-causal involved that plays an important explanatory role. This leads to the second problem: there is the suspicion that a non-causal feature of the agent’s action is doing the explanatory work.

There is, at least, one way to explain action that differs from the psychologistic approach. Reason-citing explanations (Alvarez, 2009, 2010; Anscombe, 1957; Dancy, 2002, 2018) refer to an action’s justifying reasons that render the action intelligible from the agent’s perspective. A reason-citing explanation gives an answer to the question what the agent saw in the action as choice worthy. For instance, John went to the hospital, because his father was hospitalized. Unlike as with psychologistic approaches, the explanatory power lies in facts about the world that offer reasons for certain actions. In this case, the fact that John’s father is in the hospital provides a good reason for John to go there. While John must have some epistemic relation to this reason, such as believing, the actual explanatory force is attributed to the reason itself, not to John’s psychological attitude. As Dancy puts it, “psychological states of the agent are the wrong sort of thing to be a good reason” (p. 106). In cases where it seems like the desiring is explaining actions, according to Dancy, it is actually the underlying reasons that support the desire that are doing the explanatory work (also, see Raz, 1998). So, according to a reasons-citing explanation, if John desires that p, then it is p, the reason, and not John’s desiring that does the explaining.

5.2 Explaining suicide with reasons for living and reasons for dying

While psychological suicide research often focusses on causal thinking, clinical psychologists who work with suicidal patients take a different stance through the Motivational Interviewing (MI et al., 2012) approach. I think that MI is a reasonable way to assess an individual’s Normic suicide risk, as it will become clearer in the next sections. When MI is applied to suicidality as a therapy approach, clinical psychologists focus on the agent’s reasons for living and dying. Peter Britton and colleagues state “MI is a clinical approach that currently lacks an underlying theory to explain the mechanisms by which it functions” (Britton et al., 2011). Then, they give a causal theory of why MI works, which, I think, does not take MI at face value. MI is not about mental states but about the agent’s reasons.

If a mental health professional uses MI to treat suicidality, she discusses the reasons for living and dying with a suicidal person. The clinician tries to understand why the patient sees suicide as choice worthy by understanding the agent’s reasons in favor of self-killing. Also, the clinician has to understand why the agent has not died by suicide yet by investigating the agent’s reasons for living. In some sense, the agent, having reasons for self-killing and reasons for staying alive, is in the state of suicidal ambivalence. The clinician tries to solve the ambivalence in favor of the decision to stay alive through (a) reveal features of the agent’s situation that speak in favor of living, and (b) establishing an emotional connection to reasons for living.

During a motivational interview a Vietnam veteran explains: “I have nightmares about Vietnam sometimes, and I started drinking again. I just can’t take it anymore […] I’m just tired of it all. I remember horrible things that I can’t talk to anyone about because I’m afraid of what they would think of me. And I can barely walk, and there is nothing that can be done about it.” (Britton et al., 2011, pp. 20–21) The person considering suicide is facing practical problems: chronic back and knee pain, trauma from the Vietnam War, and alcohol issues. After consulting various physicians and psychologists, he feels at a loss; nothing seems to provide relief. He then mentions that his love for his children is what keeps him alive. This not only indicates that the veteran is considering reasons for living and reasons for dying, but also highlights the severity of his practical problems. In his view, suicide appears as one potential way to cope with these issues.

Some individuals view self-killing as the best solution to a practical problem. While facing such problems, individuals consider various actions, including self-killing, in light of their goals, preferences, and values. However, it’s important to acknowledge that individuals can err in their reasoning processes. Particularly in the context of suicide, the field of clinical psychology points out potential cognitive impairments in suicidal individuals. For instance, some highlight reduced problem-solving abilities (Pollock & Williams, 1998), limited access to memory (Williams et al., 1996), or attention biases (Richard-Devantoy et al., 2016). While the process of practical reasoning for solving a problem might be biased or wrong, the practical reasons for which an agent acts make an action intelligible. If, for instance, a person kills herself because she was falsely convinced that she was suffering from terminal illness, knowing the person’s motivational reasons still makes the act intelligible.Footnote 5 We knew what the agent saw in the action.

Prospective studies indicated that, statistically speaking, participants with few reasons for living were more vulnerable for developing suicidal thoughts (Zhang et al., 2011) and attempting suicide (Galfalvy et al., 2006). Also, Gutierrez and colleagues (2000) argued that to understand suicide risk, reasons for living and dying are essential. For example, a person self-reporting a high suicide risk but who has also a strong sense of responsibility towards their family has a different risk than without having such a sense of responsibility. However, as usual in the field of suicidology, there are also studies that contradict the relevance of reasons for living. For instance, a study from Juliane Brüdern and colleagues (2018) showed that reasons for living did not function as protective against suicide. In contrast, reasons for dying were more relevant.

The relevance that Brüdern has in mind, however, depends on the predictive value of reasons for living and reasons for dying. From the perspective of traditional suicide risk assessment, reasons for living and reasons for dying are not useful, as actual suicidal behavior cannot be predicted on that basis. For developing a new model of suicide risk, I conceive different relevant aspects of reasons for living and dying in the following.

5.3 Practical rationality as a normative dimension for suicide risk-assessment

In contrast to focusing on predictive powers, the Normic account of risk considers how normal an event is, given a normative dimension. I suggest that the relevant norms are those of practical rationality.Footnote 6 Therefore, regardless of the predictive power of reasons for living or dying, they can classify an event as normal or abnormal. Accordingly, if a person self-reports strong reasons for dying and virtually no reasons for living, then the act of self-killing is, from the perspective of practical rationality, normal, as it aligns with the individual’s reasons. The criterion for normalcy here is that there’s no need for an extensive explanatory framework to understand the action’s occurrence. In contrast, if a person with strong reasons for living commits suicide, this is an abnormal event. Its abnormal nature is indicated by the considerable explanatory effort required to make the action intelligible. The latter case has to be causally explained in a complicated way by referring to a mental illness and its profile of cognitive or affective dysfunctions.

While some philosophers argue that suicide cannot be rational (Cowley, 2006; Devine, 1978), most maintain that it can be (Battin, 1995, 2005; Benatar, 2020; Cholbi, 2011; Cosculluela, 1995). For instance, according to Battin (1998), a suicide is rational “if a person reasons clearly, has a realistic view of the world and of his or her situation in it, and has adequate, undistorted information; and suicide can be a rational thing to do if it avoids harm and is in accord with one’s fundamental interests” (p. 21).

Practical reasons make an agent’s action intelligible, regardless of whether an intentional self-killing was all-things-considered rational or not. When theorists want to explain an intentional action, they can refer to the agent’s reasons to make sense of it (see Sect. 5.1). If theorists know the practical reasons for which an agent acted, it comes clear that “from the agent’s point of view there was, when he acted, something to be said for the action” (Davidson, 1963, p. 691). This way of making sense of an action is typically considered to be a rationalization (Anscombe, 1957; Davidson, 1980; Mele, 1992). For explaining the action, the justification, which rationalizes the action, becomes especially important. The relevant justification is the agent’s consideration for which the agent acted.

It is tempting to conceptualize reasons for living and reasons for dying as items on a list, with pro-reasons on the left and con-reasons on the right. Each reason would carry a certain weight, and depending on which side holds more value, one would lean more towards continuing to live or not. However, I believe this approach is overly simplistic. It is doubtful whether the practical question of whether life is worth living can be adequately addressed with just a list of reasons in favor of or against ending one’s life.

The psychologist and holocaust-survivor Victor Frankl illustrated in his book Trotzdem ja zum Leben sagen (Man’s Search For Meaning: An Introduction to Logotherapy, first published 1946; 1998) that humans have the capacity to endure unimaginable suffering if they can find a single reason to keep living. During his harrowing experience in a concentration camp under Nazi terror, Frankl maintained his will to survive by reminding himself of his reasons for staying alive, which included his determination to analyze the psychology of those enduring the worst circumstances and to publish his findings once the Nazi terror ended. Frankl’s account emphasizes that the loss of all reasons for living often goes hand in hand with thoughts of self-killing. Conversely, he observed that individuals who clung to a reason for living were able to persevere despite enduring unimaginable suffering. I think, this is where Edwin Shneidman’s approach to suicide falls short, as he presupposes that suffering as such becomes unbearable (psychache) and, then, causes suicide. In reality, suffering becomes truly intolerable when individuals lack reasons to endure it. Humans can endure immense suffering if they understand, why it is worthwhile to persist. Therefore, I think, theorist have to put special emphasis on reasons for living, namely life-sustaining reasons, when assessing suicide risk.

Life-sustaining reasons make live worth living and vary from person to person (Lopez & Snyder, 2011). Positive psychology, the psychological approach to happiness, well-being and mental health, suggests some factors that many people consider to be important for making a life worth living. Personal relationships provide emotional support and a sense of belonging. Health, being free from illness and suffering, allows individuals to engage in different activities. Accomplishing personal goals and achieving success, whether in education, career, or hobbies, provides a sense of fulfillment and satisfaction. For many individuals expressing themselves through creative work, such as art, music, or writing, gives joy and meaning. Also, helping others brings a sense of fulfillment and satisfaction. Note that being afraid of death is no life-sustaining reason. It may serve as a reason against self-killing, but it does not speak in favor of being alive.

As Frankl suggested, even if reasons to die are powerful, they are transcended by life-sustaining reasons, which are substantial reasons for living. I argue that weak reasons for living are more significant than strong reasons for dying. Life-sustaining reasons form the core of a person’s identity. It is through this core that all other reasons are weighted and contextualized. There is nothing equivalent to reasons for dying at the core of a person’s practical identity.

When considering what makes life worth living, Bernard Williams invites us to imagine a person who is contemplating suicide:

a man [who considers suicide] might consider what lay before him, and decide whether he did or did not want to undergo it. If he does decide to undergo it, then some desire propels him on into the future, and that desire at least is not one that operates conditionally on his being alive, since it itself resolves the question of whether he is going to be alive. He has an unconditional, or (as I shall say) a categorical desire. The man who seriously calculates about suicide and rejects it, only just has such a desire, perhaps. […] it [going on with life] requires that some of one’s desires should be fully categorical, and one’s existence itself wanted as something necessary to them (B. Williams, 1973, pp. 85–86).

Williams (1973) differentiates between what he terms “conditional desires” and “categorical desires.” This distinction is based on the nature of their propositional content and the requirements for their fulfillment (Connie, 2015). A conditional desire, such as the desire to complete writing a book on the condition of being alive, can be fulfilled in two ways: either by actually finishing the book (making the consequent true) or by not being alive anymore (making the antecedent false). In contrast, a categorical desire to finish writing a book can only be satisfied by the actual completion of the book, which necessitates remaining alive long enough to do so. Categorical desires, in contrast to conditional desires, both motivate a person to continue living and give her reason to live. Conditional desires only motivate an agent to do what it takes to reach a goal, but conditional desires to not make an extended existence desirable.

For instance, one person desires to get to work tomorrow. Therefore, the desire to take the train is useful, given that tomorrow the person is still alive and has a job. This is a conditional desire, because taking the train is desired on the condition of being alive. Now, consider another individual who desires the realization of an important project in life, such as having a family. Here, the individual hopes to stay alive to do this (Belshaw, 2020). For this categorical desire, the desire is only fulfilled if the agent lives long enough to make progress on the life project.

Bernard Williams introduced the idea of ground projects, in addition categorical desires, in his book Moral Luck (1981):

we need only the idea of a man’s ground projects providing the motive force which propels him into the future, and gives him a reason for living. […] in general a man does not have one separable project which plays this ground role: rather, there is a nexus of projects, related to his conditions of life, and it would be the loss of all or most of them that would remove meaning (Williams, 1981, p. 13).

For Williams, ground projects are an agent’s overarching commitments, life projects that a person identifies as central to their existence. Ground projects are essential for an individual’s sense of purpose and identity (Williams, 1981, p. 14): Identity, because a person’s practical identity is constituted through ground projects and categorical desires; Purpose, because ground projects and categorical desires make an individual’s life worth living. An agent’s ground projects reflect what they fundamentally care about, and their categorical desires are the motivations driving these concerns.

What, then, is a reason for living in this context? From a Humean perspective, desires function as reasons. Therefore, categorical desires could serve as reasons for living. One challenge to the Humean approach is that merely having a desire seems to lack a normative dimension necessary for justification. If an action is not justified from the agent’s point of view, then the action makes no sense. In this regard, theorists cannot explain intentional action. Facts are justificatory only when there is some normative dimension involved. A brute psychological fact, such as having a desire, does not make an action choice-worthy (Quinn, 1994). Consider someone with a bizarre desire to turn on the radio whenever they see one. This desire, lacking a normative dimension, is simply a behavioral disposition to turn on radios. The desire does not provide an intelligible reason for the action. Why does the agent turn on the radio? The desire alone does not contribute to an explanatory answer that clarifies what the agent sees as choice-worthy in the act of turning on radios.

Although desires may not be reasons in themselves, they can indicate an agent’s reasons. For instance, a desire can target something that the agent considers good (Raz, 1998). Here, the justifying role is played by the desired value, not the psychological attitude of desiring. An action becomes intelligible only if we understand why an agent sees something as choice-worthy. For agents, ground projects are intrinsically valuable; they are good. In this way, ground projects function as life-sustaining reasons, while categorical desires target those ground projects.

While life-sustaining reasons and categorical desires make agents carry on with their lives, they constitute a person’s practical identity simultaneously. An agent may have reasons for dying, but given the agent’s ground projects and the corresponding categorical desires, reasons for dying are fading. For a fact to become a reason for an agent, it must be relevant to their goals, desires, beliefs, or values. A fact in isolation is neutral; it becomes a reason when it connects meaningfully with what the agent cares about or aims to achieve. For example, the fact that it is raining becomes a reason to carry an umbrella if the agent values staying dry.

Reasons for dying, when there are life-sustaining reasons present, hold less relevance for an agent. They do not align with the core of the agent’s practical identity. Reasons for dying are considered through the lens of the agent’s life-sustaining reasons. These life-sustaining reasons dictate the significance of reasons for dying to an agent. As long as life-sustaining reasons exist, there remains the agent’s hope to stay alive to work on ground projects.Footnote 7

Consider an environmental activist whose life-sustaining reason is the protection of natural ecosystems. A report is released showing a significant increase in deforestation in a particular region, threatening numerous species. For many, this fact might be a reason to give up, as the situation seems hopeless. Instead of seeing this fact as a deterrent, the activist views it as a challenge to be met. The fact, the increase in deforestation, aligns with their life-sustaining reason, transforming what is a threat into a call to action. It strengthens their resolve to fight against deforestation and becomes a reason to mobilize resources, raise awareness, and engage in conservation efforts. Instead of being a reason to retreat, the fact turns into a reason to engage more deeply with the activist’s goals. Thereby, an agent’s life-sustaining reasons are critical in determining how facts are interpreted and acted upon. They can turn neutral or even negative facts into compelling reasons for action, often transforming challenges into opportunities to affirm and realize ground projects.

I want to end this section with three remarks. First, when considering the phenomenon of suicide, emotions such as sadness, anger, or loneliness seem to play an important role. Emotions influence how agents perceive reasons. For instance, fear makes threats salient, turning them into reasons for avoiding danger. Additionally, emotions involve appraisals, which can reveal reasons for action. For example, if an agent is discriminated against, anger may reveal injustice practices as a reason for action. From a reasons-citing perspective, however, reasons have the explanatory power. Primarily, then, emotions indicate reasons. If an agent feels gratitude towards someone, then the fact that explains this gratitude, such as being helped in an emergency, is the reason that explains why the agent thanks the admired person (presuming that the agent knows why they feel gratitude and why it is choice-worthy to thank the admired person). While emotions have a strong motivational component, when we explain actions from a reasons-citing perspective, we are not asking for the causing mental states, such as sadness or anger. Instead, what explains actions in the relevant sense—what makes the action choice-worthy—are the reasons that speak in favor of an action (see Sect. 5.1). It is true that there is no action without a proper motivation, but the causing motivations do not necessarily have the relevant explanatory powers.Footnote 8

Second, self-deception might be a problem for the suggested approach if individuals self-report reasons for living, while these reasons are not their actual reasons for living. Alfred Mele’s (2000) conception of self-deception highlights how agents can be driven by motivational causes to hold false beliefs about their reasons for decisions, overlooking the ‘real reasons’ behind their actions. Arguably, a study by Nisbett and Wilson (1977) illustrates how agents may develop false beliefs about their reasons for decisions. Participants were shown identical pairs of stockings and asked to choose one. Many selected the pair on the right due to a bias, yet justified their choice by claiming those stockings were of superior quality—a false belief, as all pairs were identical. The actual influence was the positioning, not the quality. This example parallels how individuals might deceive themselves about the ‘real ‘reasons for living, because the ‘real reasons’ could be, for instance, undermining their self-image.

Sandis (2015) cautions against hastily determining what constitutes an agent’s ‘real reason’. Even if based on false beliefs, these apparent reasons drive decisions. While participants are unaware of the explanatory reasons for their beliefs, if it were not for the participants’ belief that the stockings were of better quality, then they would not have chosen that pair of stockings. Therefore, I believe, as long as the agent is not lying, self-reported reasons are the relevant reasons for the agent, because these reasons actually possess the explanatory power that a reason-citing explanation approach requires.

Third, when the suicide risk of individuals is assessed based on their reasons, it constitutes a qualitative approach to risk evaluation. This does not offer the same level of objectivity as a quantitative measurement of a mental state, such as measuring hopelessness on a scale (Beck et al., 1974). Whether a fact speaks in favor of an action and functions as a reason for dying depends on the context in which an agent makes a decision, including the cultural and social context (Kral, 2012). Arguably, assessing someone’s reasons for living or dying relies on a shared understanding of the suicidal agent’s background. People from Western countries may overestimate certain reasons for living or dying for individuals with different cultural backgrounds. In this manuscript, I can only give an initial outline of the Normic account of suicide risk. A complete account of the assessment of the quality of reasons for suicide must include a conceptualization of ‘good’ or ‘strong’ that is broad enough to qualify a variety of reasons. Additionally, the account must consider the values an agent acknowledges, which is crucial for a fact speaking in favor of an action. However, I think that once an agent’s background and values are taken into account, there is intersubjectivity for evaluating reasons for action, although not objectivity in a strict sense.

6 Conclusion and outlook

From the Normic account of suicide risk, an agent’s life-sustaining reasons determine whether they have a high or low risk for suicide. This perspective differs significantly from the traditional approach. Traditionally, clinical psychologists assess suicide risk in terms of risk factors such as mental illness, previous suicide attempts, family history of suicide, and substance abuse. Thereby, suicide is not attributed to the suicidal agents and their practical reasoning. In contrast, this perspective originates from the pathological view of the phenomenon of intentional self-killing. Certainly, there are instances where suicidal action is best attributed to mental illnesses like bipolar disorder or depression. However, there are also cases where agents decide for themselves that their life is not worth living. Such decisions are based on the agent’s personal values, goals, and beliefs, and in some cases, it is presumptuous to assume that suicidal agents are irrational simply because they view their life as a burden to themselves.Footnote 9

For clinical psychology, ensuring that suicidal individuals are capable of making autonomous decisions is of utmost importance. Some individuals’ cognitive, affective, and voluntary capacities may be severely impaired due to mental illness, for example. Additionally, some agents have relevant false beliefs that undermine the action’s autonomy because the false belief prevents the action from expressing what the agent values. Psychologists should ensure that these individuals attain a level of autonomy that enables them to make responsible choices for themselves. From there, if theorists view suicidal actions as the agents’ intentional actions, then, I believe, we must respect the agents’ autonomous decision-making process. This respect includes discussing reasons and offering suggestions for solving practical problems that suicidal individuals face. Sharing viewpoints and information is not a paternalistic intervention.

The Normic account strips away suicide risk factors, such as a former suicide attempt, single marital status, substance abuse, and focusses on the agent’s reasons. From this point of view, it does not matter whether somebody has depression or not. What is important are the agent’s reasons for living and reasons for dying. At first, this might appear as risky and inappropriate, but it is a reasonable response to the inability to make useful predictions for whether an individual will show suicidal behavior or not (see Sect. 2).

If psychologists are determining who is especially in need of support, they focus on high-risk patients. Traditionally, the underlying idea was to prevent self-killing in those who are likely to act suicidal. Since this probability assessment is out of reach, an alternative approach is to consider who has the most reason to take their own life. Assuming that intentional self-killing can be an expression of an individual’s autonomous choice, those who believe they have no reason to live are the ones in need of immediate help. However, once individuals can make autonomous decisions, help cannot and should not be forced on anyone. Firstly, psychologists should ascertain that the individual is capable of making autonomous choices. Secondly, psychologists should discuss practical problems and potential solutions with suicidal individuals. This approach to treatment was already suggested by Shneidman (1998), who emphasized that the focus should be on the whole person, not just on mental disorders, and is already practiced through Motivational Interviewing (see Sect. 5.2).

This is the introduction and an initial outline of the Normic account of suicide risk, which leaves many questions unanswered. For example, the role of reasons for dying is debatable. If someone has strong reasons for dying, arguably, they are at high risk for suicide. This is because if an agent has strong reasons for dying, acting on such reasons is not difficult to explain. However, I view life-sustaining reasons as more crucial, as they form a core part of our practical identity through which we evaluate reasons for dying. Nonetheless, I acknowledge that this is debatable.