Reciprocity and epistemicity: On the (proto)social and cross-cultural ‘value’ of information transmission

: Reciprocity is a (proto)social mechanism that involves (im)politeness as a balance of positive and negative actions among individuals: doing something good to someone is expected to be reciprocated in kind (cf. Culpeper & Tantucci 2021). The same applies for negatively charged behaviour (Ibid.). The present study advances the theory of reciprocity both empirically and theoretically, as it extends the model to contexts of information transmission, i.e. cases where some news is being communicated from one interlocutor to another. What we found is that the way people react to ‘being informed of something’ remarkably involves (im)politeness and is mediated by two maxims of epistemic reciprocity: Engagement E ( be interested ) maxim and Knowledge exchange K e maxim ( be interesting in return ). Our case study is centred on Chinese telephone conversations among family members and shows that the costs and benefits realised by an information giver are matched by the information receiver when a propositional contribution to the current flow of information is produced in return. Conversely, when responses occur via bare backchanneling or absence of informative contribution to the on-going interaction, then reciprocity is not properly maintained, and perceptions of impoliteness are more likely to arise. Despite the context-dependent nature of our data, we will further argue that this finding has cross-cultural significance. Our methods triangulate between Likert-scale judgments, large scale corpus-based analysis and multivariate conditional inference tree modelling (Levshina 2015; Tantucci 2021).


Introduction
Reciprocity is a mechanism of (proto)social payments that involves the balancing of positive and negative actions among individuals: doing something good to a social persona is socially expected to be reciprocated in kind (cf. Culpeper & Tantucci 2021). By the same token, negative behaviour is 'normally 'expected to lead to a negative reaction (Ibid.). Reciprocity is arguably (proto)social as it involves responsive tendencies of human behaviour as such (e.g. Von Rohr et al. 2011), which can be identified at any stage of cultural development and in any cultural system of values (cf. Culpeper et al. forthcoming). This is not to say that reciprocity is not affected by intra-and inter-cultural specificity and, therefore, by power and distance. These are all factors that often interrupt contextually a universal tendency towards a balance of behavioural payments.
The present study is centred on epistemic reciprocity, i.e. a balance of social payments involving speech acts of information transmission, and will argue that the way people react to new information involves (im)politeness. In particular, we will show that, in the context of Chinese telephone conversations among family members, the costs and benefits realised by an information giver are perceived to be matched by the information receiver when a propositionally full and engaging contribution to the current flow of information is produced in return. On the other hand, reciprocity impoliteness is more likely to arise with silence, mere acknowledgment of what was said or absence of knowledge exchange as part of the on-going interaction. These are cases where the information receiver 'does less than expected' in return to receiving a piece of information. We will propose that two fundamental maxims determine the repayment of costs and benefits needed to maintain epistemic reciprocity: the Engagement E maxim (be interested) and Knowledge exchange Ke maxim (be interesting in return).
The research questions of this study are restricted to the contextual environment of telephone conversations among Chinese family members. However in section 5 we will argue that they can provide a baseline for cross-linguistic speculations. Our research questions are respectively: -How do people enact reciprocity in contexts of information transmission? -How do information recipients (IR) reciprocate the costs and benefits produced by information givers (IG)? -To what degree epistemic status, culture specificity and illocutionary force affect the reciprocity of information transmission? -Is it possible to identify some interactional maxims that determine the balance of epistemic reciprocity?
The present paper is structured as follows: Section 2 introduces the notion of reciprocity. Section 3 discusses the concept of information transmission from a pragmatic and ecological perspective and in relation to social reciprocity. Section 4 is devoted to the data retrieval and the results of this study, both derived from a Likert-scale questionnaire and a large-scale corpus-based enquiry. In section 5 we formulate respectively the E and Ke maxims of epistemic reciprocity. Our conclusions are given in section 6.

Reciprocity
Despite having played an important role in social and evolutionary psychology (i.a. Trivers 1971;Alexander 1980;Cosmides & Tooby 1992;Gintis et al 2008;Barclay 2012), reciprocity has been somewhat neglected in pragmatic (im)politeness theory. In Brown & Levinson's (1987) cooperation and mutual self-interest are discussed with reference to interlocutors' "mutual vulnerability of face" (1987: 61). Reciprocity is discussed more explicitly, though briefly, in Spencer-Oatey's (e.g. 2008) rapport management model, especially with regard to the concept of individuals' "sociality rights", where it is stated that "the belief that costs and benefits should be 'fair 'and kept roughly in balance through the principle of reciprocity" (Spencer-Oatey, 2002: 540). More focus to the notion is given in Ohashi (2008Ohashi ( , 2010, who proposes a "norm of reciprocity" underpinning a "debit-credit equilibrium", which is illustrated in the context of speech acts of thanking. In Ohashi's model, politeness is treated metaphorically as money, as a kind of social payment (see also e.g. Werkhofer, [1992Werkhofer, [ ] 2005Watts, 2003: 115). Culpeper (2011: 203-7) discusses the relationship between reciprocity and (im)politeness by noting that individuals establish (im)politeness thresholds involving varying degrees of politeness or impoliteness. These thresholds are updated as the interaction unfolds, via linguistic acts that might match or mismatch one interactant's behaviour. Tantucci et al. (2018) note that (mis)matches in reciprocity naturally vary diachronically and crossculturally. Guydish et al. (2020) draw on communication accommodation theory (Gallois et al. 2005) and provide an account of 'the amount of talk' that is produced by participants, while imbalances exist depending on context and social roles.
In Culpeper & Tantucci (2021) reciprocity is discussed as a key interactional mechanism serving a (proto)moral balance among interactants and among social groups. They establish a Principle of (Im)politeness Reciprocity (PIR), defined as: a constraint on human interaction such that there is pressure to match the perceived or anticipated (im)politeness of other participants, thereby maintaining a balance of payments. (Culpeper & Tantucci 2021: 150) In their framework, (im)politeness reciprocity therefore underpins a balance of costs and benefits among interactants. For instance, a politely formulated request may be expected to involve 'politeness credit 'which can be reciprocated with politely formulated compliance. Conversely, an impolitely formulated insult would entail a debit, which, in turn, can be balanced with an impolitely formulated counter-insult (Ibid.). They argue that speakers make assumptions about their mutual 'balance of social payments' based on their memory of the past relevant actions of their interlocutors and the degree of (im)politeness that they perceived. This is evident also meta-linguistically, i.e. via idiomatic formulas that are present cross-culturally, such as I owe you, I'm indebted to you, you'll pay for this, I'll remember this and so on. With the PIR there is thus a key element of "pressure" for the matching of social payments, which is primarily -though not exclusively -afforded by obligations flowing from the established (or assumed) moral order of the social structure that the interactants are part of.
The PIR provides the tools for a gradient approach to (im)politeness. This is because it is not centred on absolute evaluations of social behaviours (e.g. a directive being necessarily a face threat, cf. Brown & Levinson 1987), but rather on the dynamic and context dependent perception of how much should be done by a social persona A in order to -positively or negatively -repay what was done by another persona B.

The (proto)social dimension of reciprocity
There is also a strand of research in evolutionary psychology which has been centred on the notion of reciprocity. Trivers (1971) discusses 'reciprocity-based altruism' as a mechanism that occurs when individuals cooperate by trading helpful acts. When the benefits to the recipient of altruistic acts are greater than the cost to the actor, both participants will benefit so long as the act is reciprocated sometime in the future, with emotions such as gratitude, sympathy, and guilt having evolved to regulate systems of reciprocity (Trivers 1985). In a similar vein, Alexander (1987) views the development of moral systems as systems of 'indirect -'or somewhat delayed -reciprocity. From this angle, moral rules are established to control tendencies of individuals to behave selfishly (e.g., to cheat on a social exchange). This means that, for reciprocal altruism to evolve, a socio-normative mechanism of conditional repayment of costs and benefits must be in place among members of any cooperative activity in order to detect cheaters (Cosmides & Tooby 1992). It is an empirical fact that in human moral communities individuals hold one another accountable for their obligations. They do this via reactive attitudes -such as resentment and blame -to those who have reneged in acts of socalled moral protest, to which they expect/demand an appropriate response in return (Smith 2013 On the other hand, it would be somewhat simplistic to adopt a view of reciprocity as a mechanism that uniquely unfolds as a 'delayed 'repayment. In fact, reciprocity has been shown to be also a fundamental mechanism at play throughout single speech events or even within adjacency pairs (cf. Tantucci et al. 2018;Culpeper & Tantucci 2021). With this in mind, drawing on the intersubjective continuum proposed in Tantucci (2021), a terminological distinction is in order. We therefore distinguish between immediate reciprocity, viz. what involves the repayment of costs and benefits realised during the immediate context of the here-and-now of an interaction, versus extended reciprocity, which has to do with the long-term balancing of social payments and moral obligations among the members of any social group. The present study is centred on the pragmatics of' real-time ' (mis)matches of immediate reciprocity that are realised by interlocutors when a piece of information is transmitted from an information giver (IG) to an information recipient (IR).

(Mis)matching reciprocity
Reciprocal matching of politeness is often associated with conventionalised behaviours, as in recursive interactional patterns in contexts of greetings, partings, favours and thanks, requests and compliances, assertions and acknowledgements (Culpeper & Tantucci 2021). An example could be speaker A saying good night to B at the end of an evening spent together. At that point B will be under the social pressure to match the threshold of politeness established by A and therefore reciprocate the linguistic act with the same (or a similar) formula: good night. Absence of such responsive behaviour would normally produce a mismatch in reciprocity and be normally perceived as 'markedly' impolite. This is because silence would not meet the social expectation that -in that socio-cultural context -the utterance good night and the associated speech act should be reciprocated in kind. Culpeper & Tantucci (2021)  silence in response to a greeting). Usami (2002) notes how (im)polite downshifts and upshifts inherently "reflect the power relationships between the speaker and the addressee" (2002: 204) in Japanese interaction. This has been shown to be true also for contextually situated exchanges in British English (cf. Culpeper & Tantucci 2021: 161).  1987;Nowak & Sigmund 1998;Milinski et al. 2002). Indeed, speech acts of information transmission contribute to the establishment of the image and reputation of the interlocutors in their community of speakers, with obvious repercussions on whether they can be considered socially reliable, trustful or engaging individuals. This is true for information givers (IG), but also for information recipients (IR) and the way the latter overtly express their cooperative engagement towards the information they are receiving. Most conventionalised ways to enact this are forms of backchanneling, including nodding, pragmatic markers of agreement such as the English yeah, absolutely, strategies of alignment and so on. These are extremely frequent across languages and cultures, suggesting that the very condition of 'being informed' distinctively involves IR's pressure to repay IG's costs and benefits.

Information transmission as (proto)social cooperation
From an ecological perspective, transmission of information is far from being a neutral event.
Firstly, the assessment of the validity of information may have contingent effects. I may be told by the member of my tribe called 'James' that behind the bushes ahead of me the road is safe, whereas in reality a dangerous predator is there, ready for its next meal. Related to contingent effects of representatives are also long-term effects that may affect the social image of information givers (IG): after finding out that there was indeed a predator behind those bushes, it would not be beneficial for my survival to further rely on information given by James in the future. This mechanism has been studied throughout ontogeny, as children by the age of 4 start processing the reliability of informants, socially problematise the evidential status of propositions (e.g. Author 2021; Authors 2021) and take past accuracy of informants' representatives into account (e.g. Clément et al., 2004;Birch et al 2008;Scofield & Behrend 2008). For instance, they start predicting that a dishonest informant will provide false information (Couillard & Woodward 1999) or that an incompetent informant will be less reliable (Call & Tomasello 1999;Lampinen & Smith 1995;Clément et al. 2004). What this shows is that epistemicity does not exclusively involve the truth-value of a proposition (e.g. Palmer 2001: 24), but crucially underpins (proto)social cooperation and mechanisms of 'quid pro quo' among the members of a group. The giving and receiving of information involves social acts. Politeness theories have tended to focus on requests, apologies, etc, which are indeed social acts, but they underestimate how social the 'mere' exchange of information via assertions actually is. This is where the PIR comes into play as a model that can effectively tackle the relationship between epistemicity and (im)politeness, as the focus here is on 'balance of costs and benefits' among social behaviours, rather than the avoidance of face threats (Brown & Levinson 1987) or the pursuit of harmony (e.g. Spencer-Oatey 2008).

Information transmission and factuality
There is a balance between the speakers 'desire to provide valuable information -one that can increase their social prestige as reliable members of the group -and their need to preserve their social image in case their words be proved to be inaccurate. This entails that information givers often need to 'disclaim 'their commitment to the factual nature of what they are communicating, for example via epistemic modal, or evidential strategies, that is strategies that may turn a bare assertion into conjectures, reports, evaluations and so on.
Speech acts of information transmission are traditionally categorised as representatives, as they "commit the speaker (in varying degrees) to something's being the case, to the truth of the expressed proposition" (Searle 1976:10 Faller 2002;Tantucci 2016aTantucci , 2016b, through which the speaker is not the bearer of the truth or the likelihood of the proposition, but simply provides knowledge that s/he acquired -directly or indirectly -from the external world, as in I heard that the road is safe, apparently the road is safe and so on. This taxonomy has important implications for how speakers 'take the risk' to commit to whether a piece of information is true or not (cf. Tantucci 2016b). Recall the example made earlier about James assertively committing to the truth of the proposition the road is safe. Transmitted this way, the statement is given as a factual assertion. The (proto)social effects on James' positive face of this proposition being disproved are much more detrimental than in circumstances in which he had said I think the road may be safe (evaluative) or rather via evidential strategies such as I have just been told that the road is safe (presentative). What this means is that information recipients (IR) critically assess not only the contents of a piece of information, but also how information is This is ultimately to say that the degree to which a piece of information is transmitted as true has (proto)social implications both for information givers and information recipients. All contextual and propositional conditions being equal, the more a piece of information P is geared towards factuality, the higher the (proto)social value expected to be ascribed to P. However, this will always come with a cost, namely the risk of being disproved and the damage on the information giver's reputation as a reliable member of the group. One of the main concerns of the present study in section 4.4 will be to assess the degree to which different forms of representative speech acts may correlate with (mis)matches in reciprocity among information givers and information receivers.

Information transmission throughout turns at talk
The notion of knowledge exchange has led to a strand of research on epistemicity in conversation analysis (e.g, Kamio 1997;Heritage 2012Heritage , 2013, with social identities being shaped by interlocutors ' territories of knowledge (Berger & Luckmann 1966;Garfinkel 1967;Pollner 1987;Sacks 1984). Kamio (1997) stressed that epistemic status of interactants involves not just the actual possession of information, but the right to articulate it (Drew 1991;Maynard 2003;Terasaki 2004;Stivers et al. 2011). Heritage similarly noted that epistemic access to knowledge is a relative mechanism between interlocutors such that each occupy different positions on an epistemic gradient (Heritage 2010(Heritage , 2012. A fundamental aspect of epistemic status -one that is somewhat often neglected -is contextual situatedness. This entails that projected interlocutors' knowledge is often bound to the specific sociocultural situation in which the interaction takes place. A plumber may have, situationally, a much higher epistemic status than an emeritus professor of Physics, say, when the two are discussing the reasons why there is no more hot water in the professor's house. Similarly, in some cultures assertions made by older family members may be considered as more valuable than ones made by younger ones. This is indeed one of the research questions of this study and one that will be tackled in the case studies in section 4. Sequentiality is also key for information transmission throughout turns at talk (e.g. Bavelas et al. 2017). An example of this is the turn initial Oh in British and American English, which is normally employed for the acknowledgment that what the prior speaker has conveyed was markedly informative (e.g Heritage 2013). (1) A: Yeh, hum, I've just rung to tell you, uh, the things have arrived from Barker'n Stone.

B:
Oh, oh can I come round?
(Adapted from Heritage 2013: 560) As noted by Heritage (2103), A's turn in (1) is congruent with being exclusively within her epistemic domain, as it is acknowledged as a novel one through Oh by the information recipient (Ibid.).

Information transmission and reciprocity
The Conversation Analysis literature has also referred to reciprocity, as claims are made that 'on the record' expressions of knowledgeable positions drive interactional sequences towards an ideal epistemic equilibrium (Heritage 2012: 48). However, this view is mainly driven by 'how much is known' by each interactant. What has not yet been at issue is whether giving and receiving information involve perceptions of (im)politeness. In this study, we aim to put Culpeper & Tantucci's (2021) PIR into action and assess the way information recipients' (IR) produce overt strategies to (mis)match the costs and benefits that have been realised by information givers (IG).
The balance of epistemic costs and benefits among interlocutors will be shown to be driven by engagement, on the on hand, and by knowledge exchange on the other. The former has to do with strategies aimed at overtly acknowledging that information received is relevant and therefore 'valuable'. The latter has to do with the pressure to produce novel propositional content to return to IG. In this regard, pro-active contribution to the information flow is a decisive element of the epistemic balance of interactional exchanges. Evidence shows that children with specific language impairment tend to merely rely on backchanneling or minimal responses, as they take their turn without contributing to the information flow (Fey 1986;Hadley & Rice 1991;Lahey 1988;Bruce et al. 2010). This strand of research supports the view that operators of backchanneling alone are perceived as a somewhat 'baseline 'form of interactional engagement. This is arguably in contrast with ad hoc strategies by which information recipients pro-actively contribute to the information flow.
In this regard, we will account for the dimension of knowledge exchange by focusing on whether new propositional information is generated in return by IR. Consider example (4) below from our dataset: (2) As given in (2) above, IR merely acknowledges IG's assertion (she has never been on a boat) -via the backchannel 呃 è 'hm' -yet without producing any propositional content in response to information that he has received. In the case above our hypothesis is that people would perceive IR's response as a relatively 'weak' way to reciprocate the costs and benefits produced by IG. Things would be rather different if IR would pro-actively produce some informative response in return: ( Example (3) above is also from our dataset and includes both a backchannel and a propositionally informative comment. This is a case in which we would expect IR's response to be perceived as involving a comparatively stronger effort to match the costs and benefits produced by IG. This is because in response to I don't know if I have the address here, IR goes beyond mere acknowledgement (i.e. the backchannel 哦 ò), and also creatively responds with a new piece of propositional information: next time I will see him when I go back, therefore pro-actively contributing to knowledge exchange.
Returning to the first of the two components mentioned above, engagement in this study is  In the next section we will illustrate how the dimensions of knowledge exchange and engagement have been addressed both via Likert-scale testing and multifactorial corpus-based analysis.

Data retrieval and methodology
Our data retrieval was based on the Chinese Callhome corpus 6 , which consists of 120 unscripted telephone conversations between native Chinese speakers, comprising 250,000 words. The speakers of the Callhome were all aware of being recorded and were given no specific guidelines about the topics of their conversations. Every dialogue in the corpus is characterised by a spontaneous interaction among a younger family member (e.g. a child or a nephew) calling an older one (e.g. their father, grandmother and so on). The overwhelming majority of dialogues being between parents and their children, thus our focus was on the older/younger relationship. We manually retrieved the first 1000 representative speech acts realised by younger family members informing older ones (henceforth tagged as Y>O). We then gathered a second sample of 1000 representatives with older family members informing younger ones (henceforth tagged as O>Y). The age mismatch, however, did not play a significant role in the prediction of epistemic reciprocity (cf. results in 4.3 and 4.5).
Both datasets only included cases in which the act of informing is not elicited (as in cases where 5 Classifier. 6 https://ca.talkbank.org/access/CallHome/zho.html. Last accessed 10/08/2021. someone is asked to provide information), but spontaneously realised from one interlocutor to another.
For every speech act of information transmission, we also annotated the second pair part of the information recipient (IR) in order to assess whether his/her reaction would reciprocate the costs and benefits produced by the information giver (IG). The selection of the representatives uttered by IG was based on whether they would fit the criteria for the classification of assertives, evaluatives or presentatives (see section 3.1). Cases where more than one illocutionary force could be arguably ascribed to the IG's utterance (e.g. expressive speech acts that would also include an evaluative component such as that I think that was a horrible day) were also included in our dataset and treated with reference to their representative function, i.e. evaluative, assertive or presentative.  Table 1.

Annotation scheme
Clearly, while most of the dimensions of this model were based on formal and easily replicable criteria of annotation (e.g. presence vs absence of backchannels, presence vs absence of resonating items and so on), the dependent variable of reciprocity had to be informed by native speakers' perception of IRs' efforts made to match IG's behaviour. The next section is devoted to the design of the questionnaire and the implementation of the results into our annotation framework.

Questionnaire design
The test items for our questionnaire were randomly drawn from our dataset, which included 2000 adjacency pairs of information transmission naturally occurring from the Mandarin Callhome Some effort; A lot of effort. While our aim was to capture the degree to which the informants would perceive IR's responses as impolite, we yet decided not to include this terminology in the questionnaire. We made this choice because the layperson may not be aware that ordinary information transmission may involve (im)politeness, as this indeed was one of the research aims of this study.
Additionally, in order to assess (im)politeness reciprocity, it could make sense to determine what is the threshold of (im)politeness of the first-pair part of the exchange, so that the assessment of the response could be judged as a reaction to something that was originally perceived as 'more' or 'less' polite. This is indeed what was done in Culpeper & Tantucci's (2021) study about requestive exchanges in the BNC. On the other hand, with speech acts of information transmission not referring to IR's persona, elements of degree of imposition and face threats are rarely at issue in the IG's turn.
The focus of our study was therefore on how speakers react to commonly received information and whether their responsive behaviour as such may trigger perception of (im)politeness mismatches. The questionnaire was made of 36 exchanges which were randomly drawn from our dataset. The retrieval rational was based on the conditions given in Table 2. 7 We opted for a mixed method involving a questionnaire design. In our study of metalanguage, the reason why we did not exclusively focus on speakers' ongoing comments as evidence of reciprocity imbalances is that metalanguage of (im)politeness in conversation is a relatively rare phenomenon (i.a. Culpeper 2011), i.e. one that can be studied via ad hoc corpus-based queries in a large corpus, but not quite one that can be captured statistically throughout turns at talk of specific speech events of relatively small corpora. 8 https://www.qualtrics.com. Last accessed 10/08/21.  Finally, while the actual content of the information given could clearly not be controlled for, with six items available for every condition we were able to mitigate potential cases where the propositional content could be judged as somewhat exceptional by the informants (e.g. extra-informative or 'exceptionally' irrelevant).

Questionnaire results
Based on the questionnaire results, we fitted a Chi-square test, determining whether and how the informants 'perception of IR's degree of effort made to reciprocate the costs and benefits produced by IG was associated with the six conditions mentioned above. Firstly, we observed no statistical differences in the perception of O→Y vs Y→O. The results that we report therefore include all the 16 36 test items of the questionnaire, which indeed proved highly significant (X 2 = 658.86, df = 20, pvalue < 2.2e-16). In Chi-square testing the null hypothesis is that row and column variables of a contingency table (in our case reciprocity vs response types) are independent. Statistical significance is reached when the mismatch between predicted vs expected frequencies is above chance. Such mismatches are called Pearson residuals and are visually represented in blue and red in Figure 1 below:  intuitively triggers the perception of a negative mismatch among the balance of costs and benefits between two social personas.
Returning to Figure 1, we can see that very little effort was also significantly associated with exclusive use of backchannels (Back), i.e. in cases when IRs would acknowledge information that they received, yet without producing any 'propositional' information in return. Similar to bare silence, isolated backchannels were also perceived as involving very little effort made by IR in order to reciprocate the costs and benefits produced by IGs. Example (2) in section 3.3 is an illustration of this.
However, when backchannels were combined with additional propositional information being returned to IG, the response was then perceived as neutral, as involving a balanced match between the costs and benefits realised by IG and the ones returned by IR. This is illustrated in examples (3-4), section 3.3 and results in the perception of a matching redistribution of social payments (M), as indeed, epistemic costs and benefits produced by IG are more substantially 'paid back' by IR's propositionally contentful response.
Even more interesting was the highest weight of effort ascribed to IR's reactions that would underpin knowledge exchange, yet without the presence of turn initial backchannels: Similar to the exchanges in (4-5), both (6) and (7) involve the IR contributing to the information flow and the knowledge exchange of the adjacency pair. However (6-7) do not include a conventionalised turn-initial backchanneling marker. One might argue that this is counterintuitive. After all, one less component is present in IR's response (namely, a backchannel), which may suggest that something 'less' is done in return to IG. On the other hand, absence of highly conventionalised markers of acknowledgement suggests that both (6) and (7) are produced predominantly' on the fly', without an element that 'normally 'occurs in between the two pair parts, i.e. a backchanneling marker.
Backchannels are present in 64% of responses in our dataset (i.e. 1274 cases out of 2000), which clearly indicates that in Chinese conversation, information transmission is most conventionally responded to via specific turn initial backchanneling. When this does not happen, IR's reply is structurally less 'normal' and therefore more creative, as it is produced 'ad hoc 'to contribute to the current knowledge exchange. It is also important to note that when resonance is present, the perception of reciprocity is also comparatively higher. This entails that, in addition to propositional contribution to the knowledge exchange, formal and functional alignment with what was said by IG also plays a crucial role in 'weighting 'epistemic reciprocity.
All in all, from the analysis above two factors emerge as decisive for the reciprocity of information transmission: 'ad hoc' contribution to knowledge exchange (i.e. in the form of propositional information being produced in return) and marked engagement with IG's message (i.e. in the form of resonance with what was said by IG). These two will inform the two maxims of epistemic reciprocity that we will discuss in section 5.

Corpus-based annotation
The results from section 4.1 informed the annotation scheme of reciprocity in our corpus-based study.
What Figure 1 shows is a data-driven partition of three levels of (mis)matching reciprocity, namely including downward shift (D), match (M) and upward shift (U) (cf. Culpeper and Tantucci 20201 for a similar transposition of Likert scale results into a corpus-based annotation scheme for reciprocity).
Based on the results in Figure 1, the D level included either silence or bare presence of backchannels (Back) in IR's turn at talk. The M level regards backchannels followed by comments that would include elements of resonance (Res + Back) and ones that would not (Comm + Back). The U level included propositionally full responses that would (Res) or would not include elements of resonance (Comm). The annotation of the whole dataset checked for inter-rater reliability among three different annotators. The rate of accuracy among the annotators at each stage of analysis -including variables such as Reciprocity, IF1 and IF2 -is hereby given in Cronbach's Alphas and was respectively α = .74, α =.79 and finally α = .92.

Results from a conditional inference tree analysis
Based on our multifactorial scheme (see Table 1 in section 4), we fitted a random forest model (cf. Hothorn et al. 2006;Tagliamonte & Baayen 2012) to assess how each variable of our study affects impoliteness reciprocity. The random forest model is a machine learning method for classification that is based on a number of individual decision trees (see Ho 1995 among others). Decision trees, in turn, are multivariate models in which the outcome variable (reciprocity in our case) depends on a set of statistical "decisions" that are hierarchically ranked in terms of significance (see Levshina 2021).
The simulation of 1000 trees allowed us to reliably assess the degree to which each predictor (e.g. presence of backchanneling strategies, propositional comments, resonance and so on) would affect impoliteness reciprocity among speakers. The conditional importance of variables is visualised in Random forest for the multifactorial prediction of reciprocity The cut-off value for the interpretation of Figure 2 is the absolute importance weight of the variable with the lowest score (see Levshina 2015: 298), which always varies depending on the nature of the data. In this case, the conditional weight of each parameter ranges from around 0 (Age) to a highest score of 0.19 (Backchanneling strategies, reported in the plot as 'Back'). What this tells us is already remarkable. Namely, the Age component of epistemic status is not significant for the prediction of reciprocity in contexts of information transmission. This is an important result which confirms what emerged from the Likert scale in section 4.2. Importantly, the Chinese culture is a Confucian one, whereby a strong emphasis is given to the epistemic status of interactants. Older family members are often addressed with the honorific second personal pronoun 您 nín, instead of the more common 你 nǐ, by younger family members. A higher epistemic status is normally ascribed to older family members, as disagreement and rapport challenge are normally avoided by younger ones. This is connected with the Confucian value of 孝 xiào (filial piety) which is based on an institutionalised hierarchy of the Chinese family (Zhang 2007), with Chinese parents traditionally expecting their children to obey and conform without question and act on their commands (Hsu 1998(Hsu , 2002Wu 1996). The fact that age does not play a significant role here may suggest that reciprocity in contexts of information transmission is less affected by power imbalances than in other speech acts (cf. What is also crucial is that the three most important predictors from Figure 2 are the ones that most directly affect knowledge exchange and engagement, namely presence of Backchannels, Comments and Resonance. On the other hand, disagreement and illocutionary force appear to be very weak predictors of reciprocity imbalances. In the case of illocutionary force, that is due to the fact that 1782 out of 2000 of IG's utterances are assertions, which leads to low importance being ascribed to speech acts' variation. A conditional inference tree representation of the 'statistical decisions' leading the results above can be visualised in Figure 3: Conditional inference tree for the multifactorial prediction of reciprocity The plot in Figure 3 is entirely data-driven and is coded via the 'ctree 'function of the 'party' R package. Conditional dependencies among variables derive from statistical significance (the higher the node, the more significant the partition of each split). The top-down order of every node indicates a significant condition for assessing whether IR's response involves a downward shift (D) a match (M) or an upward shift (U). A useful way to interpret the plot above is in terms of conditional decisions made in order to predict reciprocity. In this case, the highest (and therefore most significant) split is triggered by presence vs absence of backchannels in IRs' turn initial position. Simply put, when IRs do not resort to backchannels, there is a highly significant tendency towards upward shift mismatches (U), i.e. cases where a distinctive effort is made to reciprocate the costs and benefits produced by IGs. This is indicated in the bar plot at the bottom-left corner (node 2), where U occupies around 95% of cases when IR avoids using backchannels. As discussed previously, this is most likely due to IRs' recognisable efforts to produce a creative 'ad hoc' response to what they heard. This is in sharp contrast to resorting to conventionalised backchanneling markers, as a less creative contribution to the interaction is produced in return. This is even more evident in node 4 (at the centre of the plot),  Figure 3 also shows that propositional contribution to knowledge exchange counterbalances this and determines a matching reciprocity (M) rather than a downward mismatch (D). This is particularly evident in node 5 (bottom-right) in contrast to node 4 (center).
All in all, 5 key results clearly emerge from both models in Figures 2 and 3: i. The socio-cultural component of epistemic status of age does not affect speakers' interactional tendencies towards epistemic reciprocity. This confirms the questionnaire results in section 4.3.
ii. The three most important dimensions of epistemic reciprocity are backchanneling strategies, presence of contentful comments and resonance.
iii. These three predictors indicate that a negative mismatch in reciprocity is likely to arise when information given is simply acknowledged, without some new propositional information being produced in return.
iv. Negative mismatches in reciprocity correlate with lack of overt engagement (e.g. lack of resonance) and lack of knowledge exchange (i.e. lack of propositional comments in return for what has been said).
v. Positive mismatches in reciprocity correlate with the absence of highly conventionalised backchanneling strategies. This suggests that responses that are creatively produced 'on the fly' (cf. Tantucci & Di Cristofaro ) are more likely to have a positive correlation with politeness.

Epistemic reciprocity and illocutionary force
In this section we finally aim to represent holistically the degree and distribution of reciprocity across all the 2000 adjacency pairs of our study. While illocutionary force was not a significant predictor in combination with the other covariants, that was due to disproportionally high distribution of assertions. Such disproportion in our case is important for capturing the density of interactional patterns of epistemic reciprocity that involve factual information. This is illustrated in the network graph below: Figure 5.
Networks of reciprocal (mis)matching among IG and IR's speech acts IG IR The ties connecting the bubbles in Figure 5 above represent all the adjacency pairs of this study. They show the degree of reciprocity that IRs' produced as a result of IG's representative speech acts. The latter are labeled as '←Presentatives', '←Evaluations' and '←Assertions' and correspond to the three grey bubbles at the top right corner of each graph. All the remaining bubbles appearing in light grey at the left hand side stand for IR's speech acts that have been made in response to IG. For each speech act, the bigger the bubble, the higher the frequency. Finally -and most crucially -degrees of reciprocity are expressed by the colour of each tie: blue expressing a downward shift (D), yellow a match (M) and red an upward shift (U). Figure 5 clearly shows how assertions that are responded to with acknowledgements normally lead to a downward shift (hence the dense blue coloured connections between ←Assertions and Acknowledgements). This is due to the fact that acknowledgements very often simply underpin the bare use of conventionalised backchanneling markers. Conversely, when ←Assertions are reciprocated with speech acts that are more likely to deliver factual and contentful information in return, the colour of the ties shifts in between being yellow (M) and red (U). This is most evident in the case of IRs' use of assertions in response to assertions made by IGs. In line with the discussion on illocutionary force and epistemicity in section 3.1, this result crucially supports the idea of a distinctively higher value ascribed to assertively factual information. On the other hand, presentative and evaluative speech acts tend to be far less frequent both in first and second pair parts, as indicated by the scarse number of ties in Figure 5.
All in all, what our data consistently show is that bare acknowledgment 'is not enough' to match the costs and benefits produced by an information giver. What is generally necessary in order to avoid a downward shift, is that some 'ad hoc 'propositional knowledge must be produced in return.

Discussion
Two fundamental component emerged in the analysis of epistemic reciprocity in our data: engagement and knowledge exchange. Speech acts of information transmission are 'valued 'the most when IR returns IG's efforts by producing some propositional content in return, on the one hand, and by resonating with what they heard, on the other. More specifically, from our study two epistemic.
Two epistemic maxims of the PIR are in action for determining 'currency' needed to reciprocate speech acts of information transmission:

Engagement (E) maxim:
Be interested: i. Overtly engage with P so as to convey that P is relevant.
ii. Avoid responding to P with information that bears no relevance to P.

Knowledge exchange (Ke) maxim:
Be interesting in return: i. Produce new propositional knowledge Q in return to P.
ii. Avoid absence of any propositional knowledge (either P or Q) in return to P. i. It is relevant enough to be worth the audience's processing effort.
ii. It is the most relevant one compatible with communicator's abilities and preferences. (Sperber & Wilson 2004: 612) When either or both E or Ke are deficient, a mismatch in reciprocity is likely to arise, which constitutes a formal indicator of potential absence of either i. or ii. and that the presumption of relevance is not satisfied. What is key here is that whilst relevance theory guides the communication of information and information processing, reciprocity is equally needed to account for the social expectation that relevant information should be 'acknowledged' through interaction, with obvious consequences for (im)politeness. The gradient dimension of reciprocity here is crucial, as the aim is not to assess what is impolite and what is polite in absolute terms, but rather to 'measure 'degrees of (im)politeness that are perceived due to (im)balances of social payments between social actors in context. This is an important characteristic of the PIR, as it allows to analyse situated perceptions of (im)politeness for speech acts that have rarely been taken into consideration in the literature, as in the present case of information transmission.

Conclusions
This paper further implemented the Principle of (Im)politeness Reciprocity PIR (Culpeper & Tantucci 2021) with a specific focus on information transmission. Both a Likert-scale questionnaire and data from the Mandarin Callhome corpus clearly indicate that epistemic reciprocity underpins a social balance of payments that is primarily driven by two maxims of information transmission, respectively the engagement E maxim (be interested) and the knowledge exchange Ke maxim (be interesting in return). This means that the costs and benefits produced by an information giver (IG) who spontaneously provides new information are positively reciprocated when elements of engagement and contentful information transmission are produced in return. Along a gradient scale of responsive behaviours, lack of engagement and/or knowledge exchange have a negative impact on the maintenance of reciprocity. Expectations of matching information transmission are arguably licensed by (proto)social cooperation, which requires as equal as possible a distribution of costs and benefits produced and received among members of the group. In the present case of representatives, only acquiring information without producing knowledge in return may be perceived as the behaviour of an 'epistemic free-rider', viz. as someone who enjoys the benefits of being granted information without reciprocating the effort with an epistemically 'valuable 'action in return.
While this model must be corroborated by further data from different languages, cultures and situation types (e.g. contexts in which information transmission is elicited rather than given spontaneously, or ones with different social distances and power conditions at play), nonetheless this case-study may still provide a baseline for some important speculations. The fact that age mismatch did not significantly inhibit the E and the Ke maxims in a strongly Confucian culture such as the Chinese one suggests that epistemic reciprocity may, to a degree, remain relatively unconstrained by culture-specific elements of epistemic status. Further research centred on epistemic reciprocity and contextual/cultural variation is needed to verify this. Finally, the paper emphasises that, as the PIR would predict, (mis)matches of (im)politeness tend to correlate with some degree of 'ad hoc' creativity, while highly conventionalised behaviour is normally associated with matching balance of social payments.