Habermas on rationality: Means,ends and communication

Introduction

Frankfurt School critical theorists have often criticized means-ends rationality – the ability to choose good means to ends. These criticisms miss the mark: they involve narrow kinds of means-ends rationality and say little about broader versions. Every sensible theory of rationality should include means-ends rationality.

Early Frankfurt School theorists, like Adorno and Horkheimer, could sidestep this response: the kind of means-ends rationality they attack is prominent and needs challenging. The situation is more serious for Jürgen Habermas. Because his predecessors offered little positive beyond their critique of means-ends rationality, Habermas developed a broader typology of rationality, including the crucial contribution of communicative rationality – roughly, the rationality of genuine discussion. But Habermas defines communicative rationality in contrast to means-ends rationality, and since he caricatures means-ends rationality, this undermines the very idea of communicative rationality. Most importantly, he depicts means-ends rationality as egocentric, unlike communicative rationality; but he does not define ‘egocentric’, and no definition supports this distinction, I show.

Drawing on Habermas’s late 1980s work, I offer an alternative distinction between communicative and means-ends rationality, using the ideas of sincerity and autonomy: the ends of genuine/sincere understanding and autonomous agreement are side-constraints on means which can be used in communicative rationality. But the same applies to means-ends rationality.

Michael Baurmann (1985) and Uwe Steinhoff (2009) have argued that communicative rationality is actually a subset of means-ends rationality. However, Baurmann’s comments are too brief and Steinhoff’s are too critical. I build on their arguments but instead depict the two forms of rationality as overlapping. Properly understood, means-ends rationality is not the monster it is often portrayed as, but a sensible part of any theory of rationality. It is even necessary in discourse ethics.

My constructive critique thus (a) enables a clearer and more workable idea of communicative rationality, (b) corrects misperceptions of means-ends rationality and (c) shows its relevance for Habermasian moral discourse and for critical theorists more generally, including those using communicative rationality to defend deliberative democracy. The relevance of these ideas for critical theorists is a recurring theme in this article.

I should thus say more about how communicative rationality has often been used, especially by critical theorists, to justify things it cannot justify. Sometimes, this simply reflects misreadings of Habermas. For example, in political theory Mark Pennington (2003) argues that markets can be communicatively rational, in that they communicate information rationally. But this is not Habermasian communicative rationality. In public policy, Clinton Andrews’s (2007) account of instrumental, strategic and communicative rationality bears little relation to Habermas’s. In international politics, Thomas Risse conflates communicative rationality and discourse, and sometimes discusses the ideal speech situation in precisely the concrete way Habermas now avoids (Risse, 2000, 32; 2004, 294–296, 303, 308). Such misunderstandings have not prevented instructive normative and empirical analysis: Pennington (2003: 725–737) makes important points about markets transmitting information more effectively than deliberative democracy, while Risse (2000: 23–28) casts new light on the rationality of Soviet/NATO negotiations. Nonetheless, some scholarship which portrays itself as following Habermas seems more Habermasish than Habermasian.

But some critical theorists follow Habermas too closely, summarizing his account of rationality without noting its flaws. Most importantly, many critical theorists use Habermas to attack bureaucratic decision-making and defend participatory/deliberative methods. This angle is found in political theory (e.g. Dryzek, 1990: 3–6, 9–21, 50–76, 113–116), planning studies (e.g. Forester, 1993: 24–30, 49–50, 68–72, 78–79) and management studies (e.g. Alvesson and Deetz, 2000: 90–93). Such arguments are important, and are a breath of democratically fresh air in what is often a stale, elitist literature. But means-ends rationality is needed in deliberative democracy, and communicative rationality is found in bureaucracy. I am not rejecting deliberative democracy – far from it. But it cannot be defended by drawing straight lines from Habermas’s different ideas of rationality to different types of politics (Blau, 2011: 52–54).

A brief note on terms. Habermas and his translators use different terms at different times, and some are unorthodox and confusing. To limit confusion, this article talks of the ability to choose good means to ends as ‘means-ends rationality’. ‘Instrumental rationality’, the more usual term, is here restricted to Habermas’s own, narrower idea: means-ends rationality in non-social situations. Means-ends rationality thus includes both instrumental rationality and ‘strategic rationality’, the social variant of what Habermas calls instrumental rationality.

The article proceeds as follows. After outlining and defending means-ends rationality (second section), the article summarizes Habermas’s pre-1990 account of rationality (third section) and discusses three gaps which have fostered caricatures of means-ends rationality among critical theorists (fourth section). I then challenge Habermas’s claim that means-ends rationality is egocentric: none of five plausible definitions of egocentrism supports his distinction between means-ends and communicative rationality (fifth section). Sincerity and autonomy, rather, are the distinguishing features (sixth section). Communicative rationality thus overlaps with means-ends rationality (seventh section). I conclude by offering three reasons why this all matters. First, these ideas are central contributions of one of the world’s great thinkers. Second, critical theorists need means-ends rationality, e.g. for deliberative democracy. Third, Habermas needs means-ends rationality, in discourse ethics (eighth section). ¹

What is means-ends rationality and why defend it?

Means-ends rationality is the ability to choose good means to ends. Consider a journey from London to Paris. To save time, one should choose a quicker trip over a longer one; to save money, a cheaper trip over a costlier one; to save the environment, a less polluting trip over a more polluting one. Means-ends rationality is a matter of degree: an individual exhibits means-ends rationality to the extent that she can choose good means to ends. But for ease of argument, this article depicts means-ends rationality in binary terms.

Everyone needs means-ends rationality. Someone who wants to hunt in the morning, fish in the afternoon, rear cattle in the evening and be a critic after dinner needs means-ends rationality to decide what to hunt, where to fish, how to rear cattle and what to criticize. You cannot reach these ends if you try to fish for cows or hunt in theatres. You cannot lead a fruitful life without means-ends rationality, unless you are staggeringly lucky. Means-ends rationality is one of the most basic forms of rationality, but while it exists in every complete theory of rationality (Nozick, 1993: 133, 176; Pauer-Studer, 2007: 76, 79), our ends ultimately matter more than our means (Rescher, 1988: vii, 1–2, 6–8). I must thus stress that I am not defending means-ends rationality as the dominant form of political rationality. Nonetheless, is it needed even in discursive rationality, as the eighth section shows.

Critical theorists rightly attack precisely this dominance: in modern society, they say, rationality is too often reduced to choosing means to pre-given ends (Horkheimer and Adorno, 1972: 36–42). But the main problem here is pursuing undesirable ends, or the dominance of certain ways of picking means; neither should be confused with means-ends rationality itself.

Related and equally misplaced complaints are that means-ends rationality is linked to such things as self-interest (Horkheimer, 2004: 3), treating other people merely as a means (Alker, 1990: 174), conservatism (Dryzek, 2006: 113), patriarchy (Alvesson and Deetz, 2000: 92), the predominance of science and technology (Marcuse, 1991: 157–163), the domination of nature (Horkheimer, 2004: 64), neo-liberal economics (Healey, 2006: 50, 53), bureaucracy (Dryzek, 1990: 4–5) and elitism (Forester, 1993: 28). These arguments are valuable, but as criticisms of means-ends rationality they fail: means-ends rationality does not have to be associated with any of the above. For example, Forester (1993: 28) states that means-ends rationality involves ‘treating participation as a source of obstruction’. This assertion fails by definition: when participation is a good means to ends, treating it as obstructive violates means-ends rationality. My article rethinks the relationship between means-ends and communicative rationality partly to avoid such caricatures and to show how we all need means-ends rationality, suitably understood.

Rather than attacking means-ends rationality, critical theorists should challenge people who claim to uphold means-ends rationality but who actually support things which are bad means to their ends. Habermas himself does not reject means-ends rationality (Heath, 2001: 13) but his flawed account partly explains the excessive criticisms of followers like John Dryzek, John Forester and Patsy Healey.

Habermas’s pre-1990 account of rationality

Habermas’s work is important partly because it offers positive answers to the overly negative critiques of his Frankfurt School predecessors, Theodor Adorno and Max Horkheimer. Habermas thinks Horkheimer and Adorno’s account of instrumental rationality explains why their work ‘oversimplif[ies] its image of modernity so astoundingly’ (PDM 112–113). An example is Horkheimer’s hyperbole about the ‘complete transformation of the world into a world of means rather than of ends’ (Horkheimer, 2004: 69). In fact, the problem is not means-ends rationality itself, merely the dominance of particular ends and the way in which for many people rationality is reduced to selecting means not ends. That revised argument, fortunately, is consistent with Adorno and Horkheimer’s broader social and political critique.

A similar shift of emphasis is trickier for Habermas, because his account of means-ends rationality underpins other parts of his theory. Most at threat is his idea of communicative rationality, a highly significant advance on earlier Frankfurt School theorizing. Habermas repeatedly contrasts communicative rationality and means-ends rationality, so much so that his followers sometimes think we must choose between the former and the latter (e.g. Forester, 1985: 49–51, though cf. Forester, 1993: 25; Dryzek, 1993: 214, though cf. Dryzek, 1992: 408–409). Either way, rethinking means-ends rationality will also entail rethinking communicative rationality, I show. First, though, we must cover Habermas’s account of rationality – and uncover its flaws.

From about 1965 to 1990, Habermas mainly discusses three types of rationality: instrumental, strategic and communicative. Habermas also calls instrumental rationality ‘cognitive-instrumental’ rationality, and sometimes talks of instrumental and strategic rationality together as ‘purposive’ and ‘subject-centered’ rationality. But as mentioned above I talk of ‘means-ends rationality’ here.

Since 1990, Habermas also discusses practical reason, which has three forms: pragmatic, ethical and moral. These radical changes have been overlooked by some critical theorists (e.g. Weber, 2005), but Habermas continues to use the pre-1990 account (TJ 13–14; OPC 310, 313–315, 325–326). And since it remains far better known than the post-1990 ideas, I focus mainly on the former.

The three basic types of rationality are characterized according to four criteria:

Aim of action: success or Understanding/Agreement

Criterion A produces two main types of rationality: means-ends rationality, aimed at success, and communicative rationality, aimed at understanding and agreement. Criterion B then gives two subsets of means-ends rationality: strategic rationality, which is social, and instrumental rationality, which is non-social; communicative rationality is always social. Criteria C and D overlap with A and B: means-ends rationality, whether strategic or instrumental, involves fixed ends and egocentrism, while communicative rationality is non-egocentric. (The place of ends in communicative rationality is less certain – indeed, criterion C is less prominent in Habermas’s writings, and I place little weight on it. But I do still address it briefly below, because Habermas sometimes discusses it and because some critical theorists emphasize it.)

Habermas’s instrumental/strategic/communicative typology has three great strengths. First, it moves beyond almost all previous theories of rationality by giving criteria for what makes a discussion rational. This is a stunning achievement. (For a partial 1930s precursor of communicative rationality, by Liang Shuming, see He, 2014: 66.) After reading Habermas, the essentially monological account of rationality in other writers (e.g. Rawls, 1999: 361–372) feels outmoded.

Second, and related, Habermas distinguishes between genuine and strategic discussion. In the former, participants seek mutual understanding and agreement, and use communicative rationality. In the latter, at least one participant simply tries to win, perhaps by manipulation or threats, and uses strategic rationality: she aims at her own success, not at genuine understanding and agreement. Habermas’s distinction is problematic, as we will see, but it has spawned fascinating and normatively important empirical research on genuine versus strategic communication, for example in international negotiations (Risse, 2000: 25–33) and in local planning (Sager, 1994: 102–106, 200–219).

Third, Habermas’s account of rationality partly underpins another key part of his social theory: the system/lifeworld distinction. I sidestep this for reasons of space, but it has generated important empirical research, e.g. on social movements (Schlosberg, 1995) and international diplomacy (Lose, 2001: 194–198).

So, Habermas’s account of rationality is important in itself and has sparked valuable research among critical theorists. Major problems remain, though. I will bypass issues like the scalar or dual-aspect nature of means-ends and communicative rationality (Fraser, 1985: 101; McCarthy, 1978: 27–28). I first address some gaps in the typology, not all of which Habermas tackles after 1990.

Gaps in Habermas’s pre-1990 typology of rationality

Although Habermas does not present his pre-1990 typology as exhaustive, his followers often treat it this way (e.g. Forester, 1985: 49–51). Yet it has several gaps which they overlook and which often lead them astray. I will concentrate on the three gaps which matter most for understanding Habermas’s caricature of means-ends rationality.

The first gap is that Habermas does not address discursive rationality, or discourse. To understand discourse we must see how it differs from communicative rationality. Communicative rationality is the rationality of reaching genuine understanding and agreement. In Habermas’s first extended example of communicative rationality, an older worker on a building-site tells a younger worker to buy him a drink (TCA2 121–123). The older worker makes implicit or explicit reference to subjective feelings (‘I’m thirsty’), empirical facts (‘a shop is nearby’) and social norms (‘on this building-site, older workers can tell younger ones what to do’). In communicative rationality, questioning these ‘validity-claims’ means probing them until the speaker does or could answer ‘yes’ to specific questions, such as ‘are you just thirsty, or do you want a drink? Do you mean this shop or that one? Can younger workers really be told what to do?’.

We move from communication into discourse when we move from what validity-claims are to whether they are right – whether a subjective claim actually is honest, whether an empirical claim actually is correct, whether a social norm actually is legitimate. Such questions, shows Habermas, implicitly appeal to resolution by ‘the force of the better argument’ in an ‘ideal speech situation’ or, as he now prefers, an ‘unlimited communication community’ (TP 19; LC 105–108; RC 235, 246, 272; TCA1 26, 42; MCCA 88, 202; PDM 323; JA 53, 57, 163–164; BFN 15–16, 161–162, 228, 322–323; TJ 86–87, 101–102, 105). (There are occasional exceptions: ST 36, 54; CES 209; ASI 90; TCA2 1–2.)

So, communicative rationality is merely about clarifying, understanding and accepting claims to honesty, truth and normative rightness; discourse is about establishing whether something is honest, true or normatively right (TCA1 302; MCCA 58–59). Rational communication can easily turn into discourse, and discourse involves communicative rationality, but the two ideas are analytically separate. Habermas is clear that discourse is rational (LC 107–108) and later calls it discursive rationality (OPC 309–311), rational discourse (e.g. PMT 102, BFN 107) or pragmatic, ethical and moral discourse, which are types of practical reason (JA 10–17, BFN 160–164).

But importantly, discourse’s omission from the pre-1990 typology has led some critical theorists to equate communicative rationality and discourse (e.g. Dryzek – see Blau, 2011: 49–50). Habermas, though, does not do so (see especially BFN 21; OPC 236; RC 235), as leading Habermas scholars recognize (e.g. Bohman and Rehg, 2014: Sections 3.1–3.2; Chambers, 1996: 95–101; Finlayson, 2005: 323; McCarthy, 1978: 288–292, 306–314, 323–327; Thomassen, 2010: 67, 69–70).

For this article, discourse matters for another reason: far from it being opposed to means-ends rationality, the two can overlap. When we make a claim about good means to ends, we implicitly or explicitly appeal to a discursive justification of these pragmatic claims, just as with ethical or moral claims (JA 10–11, 16; BFN 160–161, 164, 180, 186). Means-ends rationality is also present in moral discourse (see eighth section).

A second important gap is that there is no place for a group of people trying to find means to shared ends. Habermas partly fills this gap after 1990 (JA 10–11, 16; BFN 160–161, 164, 180, 186). This shows that means-ends rationality need not be non-social, an important and obvious point which has not stopped Habermas’s followers from claiming that means-ends rationality is opposed to participatory democratic decision-making (e.g. Dryzek, 1990: 4–5; Forester, 1993: 28). It is disappointing that such critical theorists repeat Habermas’s typology without probing its omissions.

A third gap is that there is no place for means-ends rationality without completely fixed ends. Habermas sometimes implies that means-ends rationality involves fixed ends (TRS 92; see also TCA1 285; MCCA 49), and some critical theorists attack means-ends rationality for this reason (e.g. Forester, 1993: 69–72). In fact, ends may not be fixed in means-ends rationality (which will turn out to be crucial for the eighth section’s examination of the place of means-ends rationality in discourse ethics). Habermas sometimes accepts this. For Weber, he writes, purposive rationality includes ‘not only an instrumental rationality of means, but a rationality of choice in setting ends selected in accord with values’ (TCA1 170, emphasis removed; also TCA1 172, 285), a point Habermas makes more often after 1990 (BFN 159–161, 180, 186–188; JA 2–3, 10–11, 63). (Alas, Weber’s translators err by translating zweckrational as ‘instrumentally rational’ rather than ‘goal rational’ or ‘purposively rational’, ignoring Weber’s distinction between narrowly instrumental and broader purposive rationality – see Weber, 1968: 26. This has fostered confusion in Anglophone circles.)

Consider an environmentalist trying to get from London to Paris. If there is no green route, she might cancel the trip. Her values lead her to change her immediate ends, because she recognises that this specific end will clash with her deeper ends/values. Instrumental rationality is not only about how to reach an end but about the consequences of that end for one’s other ends/values.

This important idea is usually overlooked, and Habermas himself often elides the two aspects of means-ends rationality (TCA2 303, 332–333; AR 258; OPC 219; TIO 26–27). It does not help that Habermas or his translators talk of them, respectively, as ‘instrumental rationality’ and ‘rational choice’ (TRS 91; TCA1, 170, 172; but compare CES 117). Given the social-science meaning of ‘rational choice’, this term is now unhelpful.

Means-ends rationality can thus help us choose ends if we ask whether proposed ends are themselves good means to other ends (Simon, 1983: 7–8, 11). The environmentalist above has two ends: reaching Paris and protecting the environment. Reaching Paris is a bad means for protecting the environment, which takes precedence; so the environmentalist decides not to go to Paris. The place of such reasoning in Habermasian discourse is accepted, in different terms to mine, by Rehg (1994: 47–48, 71) and Chambers (1996: 90–91).

Of course, means-ends rationality cannot itself say whether reaching Paris or protecting the environment is more important. The environmentalist used means-ends rationality to cancel the Paris trip because environmental protection was already prioritized. Means-ends rationality only helps us evaluate ends in limited ways. As discussed above, then, means-ends rationality is merely one part of a broader theory of rationality. Still, the point I am making helps illustrate Weber’s and Habermas’s recognition that means-ends rationality need not assume completely fixed ends. Ends are also being chosen – even if only in minor ways – in non-social situations.

In summary, Habermas’s criteria A, B and C do not overlap well. That is problematic both because it has led many critical theorists astray, and because it highlights Habermas’s caricaturing of means-ends rationality in ways that make it seem impermissible in discourse ethics.

Means-ends rationality and egocentrism

I now address the assumed overlap between criteria A and D. Habermas regularly depicts means-ends rationality as egocentric, and implies that actions aimed at genuine understanding and agreement cannot be egocentric. Neither is right, I argue. Again, spotting this helps us see how means-ends rationality can apply in discourse ethics.

One example of such argumentation, involving strategic rationality, is worth considering in a little detail, partly because we see Habermas smuggling in egocentrism rather revealingly, and partly because the sixth section presents strategic rationality as the key to understanding the difference between means-ends and communicative rationality.

Strategic rationality is the social version of instrumental rationality. Habermas distinguishes strategic and communicative rationality in several ways; even sympathetic commentators question his account (Johnson, 1991: 188–191). But one key distinction is egocentrism. Revealingly, Habermas twice discusses means-ends rationality without initially mentioning egocentrism – but then sneaks it in when trying to distinguish strategic from communicative rationality. He starts without a hint of egocentrism: means-ends rationality involves an actor trying to reach ‘an end’. But then we find an egocentric rabbit in Habermas’s hat: communicative rationality involves agents who ‘are not primarily oriented to their own individual successes’ and whose actions ‘are coordinated not through egocentric calculations of success but through acts of reaching understanding’ (TCA1 285–286; RC 263–264). The clear implication – that instrumental and strategic action are egocentric – is simply asserted, not justified. Nor can it be justified, as the second section argued.

More generally, Habermas’s depiction of strategic rationality errs revealingly. He uses the term ‘strategic’ because he thinks it is the rationality of game theory, whose goal is to find strategies maximizing one’s interests (TP 271–272; CES 117). But Habermas’s grasp of game theory is limited and he overlooks game theorists who model non-selfish interests (Johnson, 1991: 189–191, 199). (Habermas’s reply seems to miss Johnson’s point – BFN 554.) Habermas wrongly thinks that strategic rationality entails ‘egocentric utility calculations’ (MCCA 133–134; see also RC 237).

This last comment suggests one way of understanding egocentrism – which is important, because to my knowledge Habermas never says exactly what ‘egocentric’ means. There are five likely possibilities:

Solipsistic egocentrism – ignoring others’ interests.

Solipsistic egocentrism means that an egocentric actor ‘doesn’t consider the interests of others or recognize they differ from the actor’s, and doesn’t relate [the] two points of view’, following Kohlberg (MCCA 128). But this is not a necessary part of means-ends rationality. A parent can seek good means to satisfy her children’s interests in the full knowledge that her own personal interests are suffering. Indeed, strategic rationality as Habermas and game theorists depict it requires individuals to consider others’ interests, recognize when they differ and choose an appropriate strategy (Johnson, 1991: 190).

First-person egocentrism ² means focusing only on yourself, whereas communicative action shifts our perspectives. This might seem to fit a tricky passage where Habermas explains how in strategic rationality each participant’s orientations are ‘egocentrically geared toward the requirements of each actor involved’, i.e. focusing on ‘the actor’s own success’, but communicative action ‘interrupts’ this, ‘temporarily chang[ing] the mode of action’ to an ‘intersubjectively shared’ situation. Participants ‘must shift perspective from the objectivating attitude of an actor oriented toward success who wants to realize some purpose in the world, to the performative attitude of a speaker who wants to reach understanding with a second person with regard to something in the world’ (OPC 224; emphasis removed).

Steinhoff thinks this passage ‘alternates between triviality and nonsense’ (Steinhoff, 2009: 34). This is too strong – Habermas seems to be getting at something significant that I hope to reconceptualize in the next section. But Steinhoff is right to question the passage. Perhaps the biggest concern is that the start of the paragraph states that:

even in communicative action, the teleologically structured sequences of action of the individual actors pervade the processes of reaching understanding; it is, after all, the purposive activities of the participants in interaction that are linked up with one another via the medium of language.

Either way, the strategic rationality of game theory regularly involves seeing things from others’ perspectives: that is often what makes it strategic. And a parent acting in the interests of an unborn child will try to pick good means to those ends, again from another’s perspective. First-person egocentrism cannot be a defining feature of means-ends rationality.

Instrumentalizing egocentrism means that ‘other persons are accorded merely the status of means or limiting conditions for the realization of one’s own individual plan of action’ (JA 5–6). Whether or not Habermas is referring to egocentrism here – the text is ambiguous – this form of egocentrism must be addressed because some critical theorists do think it is what means-ends rationality involves (e.g. Alker, 1990: 174). Philosophers still dispute what treating someone merely as a means entails (Parfit, 2011: 212–232), but all that matters for us is that just as means-ends rationality may or may not be solipsistic, so too it may or may not involve treating someone merely as a means, as with parents seeking good means to their children’s ends. Despite the similar terms, treating others ‘instrumentally’ is not a necessary part of instrumental rationality. Moreover, communicative rationality can involve treating others as means to one’s ends. In Habermas’s building-site example of communicative rationality, above, an older worker who gets a younger worker to fetch him a drink could be treating the younger worker as a means.

Egotistical egocentrism is the most obvious interpretation of ‘egocentric’. Habermas equivocates here, as the next section discusses, but he often depicts means-ends rationality as self-interested. Means-ends rationality in general and strategic rationality in particular are ‘oriented only to each’s own success’ (CES 41, 117), a claim made repeatedly after 1990 (JA 5–6, 15; BFN 25, 27, 90, 161, 337; TIO 12–13, 21, 24, 25, 31, 32). He also implies that strategic rationality is egocentric while communicative rationality is not (TCA1 285–286; RC 263–264; PMT 192; OPC 224, 233).

But means-ends rationality need not be egotistic. One can clearly choose means to non-egotistic ends, as when a parent acts in the interests of an unborn child, or a consumer uses environmentally-friendly shopping methods. Both involve individuals in non-social situations acting instrumentally but not selfishly. Steinhoff (2009: 29), likewise, criticizes Habermas for ignoring altruistic means to ends. True, non-selfish motivations can have selfish elements, such as getting satisfaction from paying taxes out of duty. But that is different to self-interestedly paying taxes for fear of punishment. Rational choice theorists need not assume narrow self-interest: they need only require that individuals pursue ‘their preferences’, which may include ‘empathy for family, friends, whales, trees, or random strangers’ (Shepsle and Bonchek, 1997: 17). This point has been made in Habermasian circles (Heath, 2001: 57; Johnson, 1991: 190; see also Dryzek, 1992: 406) but some critical theorists overlook it (e.g. Risse, 2000: 3). Writers should stop linking means-ends rationality to narrow self-interest. Rational choice theorists themselves invite such confusion when they talk of ‘rational self-interest’ (e.g. Moe, 1990: 219), which conflates means-ends rationality and self-interested ends.

Non-duty egocentrism is a Kantian interpretation implied by Habermas’s post-1990 writings. ³ Kant describes three motivations for action: self-seeking, as in the shopkeeper who keeps prices low to maximize profit; inclination, as in the shopkeeper who keeps prices low out of love for his customers; and duty, as in the shopkeeper who keeps prices low because this is right and reason tells him so – the ‘good will’ (Kant, 1997: 7, 10–12).

While Habermas’s pre-1990 writings only touch on Kant (the main exception being MCCA 195–211), Habermas now places discourse ethics ‘squarely in the Kantian tradition’ (JA 1), moral discourse being a dialogical variant of Kant’s categorical imperative (JA 6–10). I am far from certain that Habermas has non-duty egocentrism in mind after 1990, but it fits what he says in the crucial section of Justification and Application where he distinguishes pragmatic, ethical and moral forms of practical reason (JA 5–8), the former being equivalent to means-ends rationality.

Unfortunately, even if it is central to discursive rationality, non-duty egocentrism cannot distinguish communicative and means-ends rationality, because communicative rationality too can be egocentric in this sense: it is the rationality of genuine understanding and agreement, and has nothing to do with duty. Habermas never says, or implies, that we are communicatively rational only when driven by duty. And his examples of flight attendants and building-site workers simply involve people trying to understand each other in order to coordinate their actions (TCA1 300–301; TCA2 121–123). Both scenarios are consistent with participants thinking in non-duty-based ways and acting for non-duty-based reasons.

In sum, neither instrumental nor strategic rationality as Habermas defines them – choosing means to ends in non-social and social situations respectively – are necessarily egocentric. And egocentrism may be present in communicative rationality. So, criterion D – egocentric or non-egocentric motivation – does not distinguish means-ends and communicative rationality.

Unless someone can find a definition of egocentrism which fits Habermas’s claims, egocentrism is a non-starter. Habermas’s position thus seems to be untenable; so too that of the many critical theorists who repeat him uncritically. Critical theorists can legitimately attack self-interested ends, but this is only one application of means-ends rationality, not a problem with means-ends rationality itself. Habermas’s stipulations about means-ends rationality’s egocentrism are unjustified and unjustifiable.

Strategic and communicative rationality: Sincerity and autonomy

Autonomy, I suggest, is a better way to distinguish means-ends and communicative rationality. In three essays from 1985 to 1990, Habermas alters the distinction between means-ends and communicative rationality (AR 240–242; OPC 217–220; RD 130–131; see also ASI 213). His comments are not always clear (see especially AR 241–242), but one set of distinctions is very instructive. For instrumental action, writes Habermas, ‘the goal … is determined (a) independently of the means of intervention (b) as a state to be brought about causally (c) in the objective world’. Communicative action differs in each respect. As regards (a), the linguistic means of reaching understanding are an inherent part of the goal of reaching agreement. ‘Grammatical utterances do not constitute instruments for reaching understanding in the same way as, for example, the operations carried out by a cook constitute means for producing enjoyable meals’. As regards (b), the speaker cannot herself cause agreement: all interlocutors must agree autonomously. I take Habermas to be saying that a cook can make an omelette but a speaker cannot make an agreement: the speaker cannot make a listener autonomously agree. This is essentially Locke’s point that beliefs by definition cannot be coerced. (See too the similar distinction given by Heath, 2001: 78–80.) As regards (c), instrumental actions occur in the objective world, communicative actions in the intersubjective world of language (OPC 217–220; AR 240–242; RD 130–131).

This is more rigorous than the earlier account. Still, distinctions (a) and (c) are overdrawn. We always risk being overly influenced by our examples, and Habermas’s examples of instrumental actions in the objective world, such as cooking, underplay choices of means in communication. A speaker must decide what to say and how to say it. If asked to justify empirical claims, for example, she should consider how best to do so – personal stories, statistical evidence, etc. Meanwhile, some instrumental choices are as essential to instrumental action as language is to communicative agreement. You cannot make an omelette without breaking eggs. You may have to make choices about how many eggs to use, who should break them, how hot to cook them, etc. Similarly, you cannot communicate without using linguistic tools (including not only language but also silence and body posture, of course), and you must choose good means of communicating your ideas – stories, statistics, signing, singing, etc. This type of means-ends rationality is also implied by Habermas’s comments about speech being ‘a medium in which … linguistic means … are employed instrumentally’ (CES 68; also RC 237). Again, my position will not surprise people who see means-ends rationality simply as the ability to choose good means to ends. (For parallel criticisms, see Heath, 2001: 51–52.)

Nonetheless, distinction (b) is extremely useful. This must be a key difference between strategic and communicative rationality. In communicative rationality, an individual makes claims about what is honest, true and right, but lets others decide autonomously. You can bring your argument to others but you cannot make them believe it. You can bribe or coerce them to accept it (strategic rationality) but only autonomous acceptance means they genuinely agree with it (communicative rationality).

Autonomy is thus the first key difference between strategic and communicative rationality. Sincerity is a second: communicative rationality requires that participants agree to sincere validity-claims. If a speaker is dishonest about subjective feelings, empirical facts or social norms, then even if a hearer autonomously accepts the claim, her understanding is not genuine, and the discussion as a whole is not communicatively rational. (Note that Habermas scholars often use ‘sincere’ for subjective validity-claims only. But I now think it is better to talk of ‘honest’ here, and use ‘sincere’ for any genuinely intended validity-claim, whether subjective, empirical or normative.)

Autonomy and sincerity are thus clear and useful distinctions between strategic and communicative rationality. Note that autonomy and sincerity may be self-interested: in Habermas’s building-site example, the older worker wants a drink or might simply like bossing younger workers around, and this does not preclude communicatively rational agreement. Note, too, that communicative rationality is compatible with hierarchy (TCA1 300–301; TCA2 121–123; see too Chambers, 1996: 95–96). Indeed, lifeworld norms can be patriarchal and repressive (Fraser, 1995: 24, 35–36).

A new view of communicative and means-ends rationality

The above arguments require significant revisions to Habermas’s distinction between means-ends and communicative rationality. Egocentrism, long upheld by Habermas and his followers as key to the distinction, is irrelevant: autonomy and sincerity are central. Communicative rationality requires autonomous agreement. Crucially, though, means-ends rationality may also require autonomous agreement. If Jack wants to marry Jill and seeks her autonomous assent, threatening her with a gun is not a good means to that end.

Similarly, communicative rationality requires sincere understanding, i.e. understanding based on sincere validity-claims. But again, means-ends rationality too may require sincere understanding. If Jack wants to marry Jill and needs her to understand that he really loves her, telling her only that their marriage would help their water-carrying business is not a good means to that end.

This implies that far from communicative and means-ends rationality being alternatives – even enemies – they overlap. This argument is not entirely original although its implications have not been fully grasped. Baurmann spends a paragraph discussing how someone seeking consensus with a discussant ‘certainly aims at an identifiable end, for which he must use suitable means’, e.g. he seeks ‘the best possible arguments … whose suitability [he] examines in an altogether success-oriented way’ (Baurmann, 1985: 189). Baurmann’s position strikes me as correct, but it is very brief, and ‘success’ here is not just ‘winning the other’s agreement’ but winning the other’s autonomous and sincere agreement.

Steinhoff goes further, depicting communicative rationality as a subset of means-ends rationality: ‘communicative action is only a particular form of strategic action’, as communicative action is ‘evidently directed at goals’ (Steinhoff, 2009: 35, 23; see also 12). What follows is too negative: Steinhoff (2009: 77) seems to imply that as communicative rationality is a subset of means-ends rationality, there is little to say about communicative rationality. Indeed, he pronounces the whole Habermasian project ‘a failure’ (Steinhoff, 2009: 242). If it is a failure, it is a glorious one, as with Aristotle, Hobbes or Marx, and one just as worthy of continued study. But I have argued that communicative rationality is not a failure: it remains an important and incredibly original insight into the rationality of discussion, especially if we refine the idea.

Moreover, just because communicative rationality is aimed at success, and just because achieving successful communication requires picking suitable means, does not mean that communicative rationality is a subset of means-ends rationality, merely that the two can overlap. Communicatively rational individuals must act in means-ends rational ways, but this does not make communicative rationality a subset of means-ends rationality, for two reasons. First, what makes a conversational partner’s claims rational is that they are subjectively honest, empirically true and normatively right; this makes them good means, but they are not good means as such. Second, a communicatively rational conversant must also pick suitable means, e.g. deciding whether to tell a story or quote statistics, but this only shows that means-ends rationality is used by communicatively rational people, not that communicative rationality is means-ends rationality.

What do the criteria of autonomy and sincerity imply for our understanding of communicative rationality? To grasp this, we must see how ends affect the choice of means in means-ends rationality. Means-ends rationality is the ability to choose good means to ends, so ends must affect which means are good: someone whose immediate end is to get to Paris, and whose more general end is environmental protecting, may pick a different means of transport to someone who has no such environmental concerns. Baurmann (1985: 191) makes a similar point.

More importantly, some ends exclude some means. Consider a modified version of Habermas’s example of how to get to work when one’s bicycle breaks (JA 2–3). Anyone whose ends are broadly liberal cannot use means such as killing someone and stealing her car, even if this is the quickest way to get to work. Such ends act as side-constraints on the choice of means. The idea of side-constraints will prove central to understanding the relationship between means-ends and communicative rationality.

I thus turn to communication. A subset of means-ends rationality is the rationality by which two or more people reach understanding and agreement. Call this ‘conversational rationality’. Conversational rationality has two subsets: strategic-conversational rationality and communicative rationality. Strategic-conversational rationality is the rationality by which one or more people try to get non-genuine and/or non-autonomous agreement from one or more other people, for example by using misinformation or threats. (Strategic-conversational rationality is only one kind of strategic rationality: a chess player out-thinking her opponent uses strategic rationality but not conversation. I must thus introduce a new term into Habermas’s typology.)

Communicative rationality is the rationality by which two or more people try to reach genuine understanding and autonomous agreement. Sincerity and autonomy are side-constraints on means: misinformation and threats are ruled out, because they preclude genuine understanding and autonomous agreement, in the same way that ends can be side-constraints in means-ends rationality more generally.

The key distinction between strategic-conversational and communicative rationality is whether validity-claims are sincere and agreement is autonomous. Strictly speaking, I should say the extent to which, not whether, validity-claims are sincere and agreement is autonomous. Habermas tends to talk in overly stark terms, as critics like Lasse Thomassen (2010: 65) and Samantha Ashenden (2014: 432–437) rightly point out. But as the second section noted, this article talks in binary terms for ease of argument.

The distinction between strategic-conversational and communicative rationality matters normatively: in politics we often want genuine understanding and autonomous agreement but actually get misinformation and threats (Dryzek, 2006: 113–123). Although I have been very critical of Dryzek’s reading of Habermas (Blau, 2011), Dryzek’s broader normative insights now strike me as far more significant than these misreadings.

So, communicative and strategic-conversational rationality are not fundamentally different. The only difference is the ends which govern the choice of means: in communicative rationality, the ends of genuine understanding and autonomous agreement are side-constraints on means.

Communicative rationality may be seen in terms of speakers, hearers, and intersubjectively. A speaker makes choices about what and how to communicate; some means are better than others for reaching genuine understanding and autonomous agreement. Hearers, meanwhile, must correctly infer what speakers mean. This helps us see how communication can be rational. Consider this example:

Larry: Would you like a coffee?

Mahika: Coffee would keep me awake.

Such inferences also arise in means-ends rationality more generally, when deciding how to cross a road, or where to pass a football to give someone a scoring opportunity. True, means-ends rationality sometimes involves conscious, complex computations, as when bureaucrats weigh up different policy options. But this can also apply in communicative rationality, for example when listeners think hard about whether they accept what was said or when writers ponder how best to communicate complex ideas.

The above account was essentially monological, involving what happens in each communicator’s head. Three intersubjective features of communication should also be addressed. First, communication requires shared understandings. For example, in Habermas’s building-site example, the older worker says ‘I’m thirsty’. Like Mahika’s response above, this statement is an ‘implicature’ (Grice, 1989: 24–40): by saying one thing, the speaker also means something else – here, ‘I’m thirsty [so get me a drink]’. The older worker implicitly draws on the hierarchical norm allowing him to order younger workers around. If the younger worker does not know this norm, he may interpret the older worker as describing his thirst rather than requesting a drink; communicative rationality is not achieved. Habermas’s ‘lifeworld’ is similar in this respect to Searle’s (1983: 143–154) ‘Background’ and Bach and Harnish’s(1979: 5) ‘mutual contextual beliefs’. Indeed, Habermas’s work on communication often coheres with analytic philosophy of language (Searle, 1991: 90; for more detail, and problems with Habermas’s philosophy of language, see Thomassen, 2010: 62–66).

Second, communicative rationality often requires hearers to ask questions clarifying what speakers mean. If the bemused younger worker asks what the older worker means, and the older worker explains the hierarchical norm, the younger worker now understands the older worker’s statement. We might say that there has been a fusion of lifeworlds.

Third, and most important, communicative rationality is inherently intersubjective, involving agreement. One person cannot make another autonomously agree. When someone simply tries to win an argument by any means necessary, this is governed by strategic-conversational rationality, not communicative rationality. Many monological processes will be the same in both kinds of rationality: for example, hearers must make correct inferences to understand what someone means. But autonomous agreement imposes extra hurdles.

Habermas might respond, following his 1988 essay, that means-ends rationality involves choosing objective means, such as which route to work is shortest, whereas autonomous acceptance of a validity-claim involves something intersubjective which a speaker cannot herself determine. But means-ends rationality can include means which require autonomous intersubjective agreement, for example if your bicycle breaks and you would not use a neighbour’s bike without her consent. Your neighbour’s bike could be a good means to your end – but not if it clashes with ends such as respecting property and consent.

Another objection to my position would be to say that communicative rationality has elements not found in means-ends rationality. Consider two people who are sincere but cannot reach autonomous agreement because the hearer makes faulty inferences and thus misunderstands the speaker. These faulty inferences are failures of rationality; but, the objection goes, they are not about choosing bad means to ends, and so communication is not an example of means-ends rationality.

I disagree. Communicative rationality requires that speakers and hearers choose good means to genuine understanding. When speakers are overly ambiguous, hearers should ask questions to clarify what is said. Clarificatory questions may thus be good means for reaching the end of genuine understanding. This again highlights the place of means-ends rationality in communicative rationality.

My account of communicative rationality fits much that Habermas says about communicative rationality, but discards most of his comments on means-ends rationality, which are inaccurate, unhelpful or inadequately distinguish means-ends and communicative rationality. My emphasis on sincerity clearly fits Habermas’s position, and while I stress autonomy more than he does, it fits his broader position and is vital for distinguishing communicative rationality and means-ends rationality. Overall, my position is broadly Habermasian, but definitely not Habermas’s, and requires us to free ourselves from the shackles of caricatured understandings of means-ends rationality.

Conclusion: So what?

I have criticized Habermas and many critical theorists for caricaturing means-ends rationality, presenting it as egocentric when this need not be so and depicting communicative rationality as non-egocentric when this can be so. Far from means-ends and communicative rationality being opposed, they overlap.

Why does this matter? I will give three reasons: one intrinsic and two instrumental. The intrinsic reason is simply that communicative rationality is a central contribution of one of the world’s great thinkers, and to the extent that we can clarify or improve this idea, so much the better.

The first of the two instrumental reasons involves critical theorists. This article has repeatedly shown that Habermas’s comments on means-ends and communicative rationality have been used (and often misused) by critical theorists. They, even more than Habermas, have tended to caricature means-ends rationality. This stance is undermined when we see overlaps between means-ends and communicative rationality. Communicative rationality can and should be defended positively, not on the back of caricatures.

This point is particularly important because critical theorists may need instrumental rationality. Consider those defending deliberative democracy, such as Dryzek in political theory and Healey in planning studies: they need means-ends rationality to work out how to implement deliberative democracy. For example, one fascinating Habermasian study shows how role-playing games foster empathy, helping participants from different backgrounds debate problems open-mindedly (Innes and Booher, 1999). This embodies means-ends rationality: it helps other Habermasian deliberative democrats work out better means to their discursive ends. If you want to implement Habermasian principles in democratic settings, you should consider using role-playing games: they might be good means to your ends.

But the second instrumental justification of my article is more significant: means-ends rationality is needed in discourse ethics. I am not just making the claim that discursive communicants need to pick forms of argumentation that will convince others, as Baurmann discusses (see above). Rather, reaching discursive conclusions about norms almost certainly requires thinking instrumentally, to some extent.

Recall Habermas’s universalization principle (U), one of the two core principles of discourse ethics:

All affected [by a norm] can accept the consequences and the side effects its general observance can be anticipated to have for the satisfaction of everyone’s interests (and these consequences are preferred to those of known alternative possibilities for regulation). (MCCA 65; emphasis removed; see also MCCA 197)

Habermas says surprisingly little about what this involves, but the focus on consequences and side-effects must entail means-ends rationality. It also entails what John Harsanyi calls ‘criterion-satisfying’ rationality: to climb the highest mountain in California one must climb Mount Whitney, and it seems odd to call this a good means to one’s end when it is constitutive of the end (Harsanyi, 1986: 85). This point is made in different terms by William Rehg. When Habermas refers to the ‘consequences’ of a norm, Rehg argues, he means two things. First, what Rehg calls the norm’s ‘action constraints’ (which is equivalent to Harsanyi’s criterion-satisfying behaviour). If two business partners are discussing a quick-profit scheme and another says ‘but wouldn’t that amount to theft?’, she is claiming that the scheme satisfies the criteria of theft and should thus be excluded. Second, there are broader consequences. For example, capitalist exchange may seem perfectly reasonable at the level of the participants but may have undesirable systemic consequences (Rehg, 1994: 46–49).

Now imagine a discourse about whether gay people should be allowed to marry. Such discourse will again involve both criterion-satisfying and means-ends elements. Discussants will inevitably make criterion-satisfying assumptions about the meaning of terms like ‘gay’ and ‘marry’, and such assumptions will likely bubble into explicit discussion and perhaps be refined. Discussants must also consider the consequences of different proposals, on gay people, bisexuals, trans people, cis people, straight people, children and so on. Discussants should even consider the ‘expressive harm’ of the various regulations: for example, does the category of ‘civil partnerships’ imply that such couples have less status than traditionally-married couples? (Criterion-satisfying and means-ends reasoning will overlap here.) Participants should consider the broader consequences of different regulations: what kind of society is being promoted and what does this in turn mean for individuals’ interests? Note also the relative component of Habermas’s theory. He does not want participants to be guilty of the Nirvana fallacy, rejecting positions which seem flawed or supporting positions which seem strong: every position must be assessed in terms of its strengths and weaknesses and contrasted with equivalent comparisons from alternative norms.

So, in all such discourses we will consider whether the means being debated are good ones for the satisfaction of interests. This involves means-ends rationality. And note that this is not egocentric. In discourse, I have to consider whether a proposal is a good means to satisfy your interests, not just mine, and you have to do the same for me.

In conclusion, Habermas’s account of rationality is significantly flawed but not fundamentally flawed. The flaws are significant: he must drop his career-long claim that means-ends rationality is egocentric, and he should accept that communicative and discursive rationality can overlap with means-ends rationality and that discourse ethics must entail means-ends rationality. But the flaws are not fundamental. Communicative rationality, appropriately reworked, survives; it significantly improves our armoury of ideas about rationality in general, and the (ir)rationality of much political discussion in particular. Critical theorists can still oppose those kinds of means-ends rationality which are egocentric, instrumentalizing or anti-participatory. Examples of such means-ends rationality are hardly uncommon in contemporary politics.

The great irony about the repeated invectives against means-ends rationality by Habermas and many of his followers is that it has been present in discourse ethics all along. If discourses are islands in the sea of communicative practice (e.g. MCCA 56), then means-ends rationality is the sand that forms part of the islands and is often carried by the sea. Habermas and his followers have caricatured means-ends rationality as egocentric, but this view has no justification – in Habermas or in theory. Far from means-ends rationality being something that Habermas and critical theorists should oppose, it is something they should welcome.