When “PO” Does Not Mean Respect: Honorific Form, Relational Conduct, and AI-Mediated Talk Among Tagalog-Speaking Filipinos

1. Introduction

Po is easy to spot, but respect isn't. For a long time, Tagalog grammars have regarded po and similar forms as part of the respectful or polite way of addressing people, along with forms such as opo, kayo, ninyo, and inyo (Schachter & Otanes, 1972; Kess, 1982). It's not that this description is wrong. The issue arises when the description is asked to do something beyond what it is capable of. The fact that po appears informs the analyst that a conventional honorific has been employed, but it doesn't tell the analyst what the speaker thinks of the person being addressed, why the form has been kept, or whether the situation surrounding the utterance is conciliatory, reluctant, irritated, formal, affectionate, or sarcastic.

The everyday use of Tagalog includes simple forms. Disagreeing with the statement "Hindi po ako sang-ayon" is still a matter of disagreement. To say "Mali po kayo" is still a way of correcting something. The expression "Opo, kayo na po ang tama" may sound like sincere acceptance, or it might indicate weary surrender or irritation, depending on the situation and how it is delivered. All of this does not mean that po is without significance; on the contrary, it shows that the form should not be made to convey the entire meaning by itself. For instance, Gocheco (2009) found that po is used in various speech acts in Philippine campaign advertisements, where it can be employed for persuasive purposes, mitigation, and in public address. Kess (1982) situated the respectful forms in Tagalog within the contexts of familiarity, equality, rank, and group membership rather than reducing them to age alone. More recent evidence from parallel corpora also indicates that po is selectively used rather than being mechanically derived from every honorific cue in another language (Jeon, 2026).

The other case is just as significant. A daughter might use Ingat, Nay without the particle po and yet still be expressing a deeply respectful relationship. Friends who are close to one another may talk without employing honorifics but will still keep to their duties of care and concern. Since the use of po cannot be taken as proof of personal respect, its omission cannot be used to show that there is no respect either. This basic symmetry is the foundation of the study, since it prevents one linguistic form from serving both as the definition of respect and as evidence of it.

The study of politeness provides a way of making this distinction. Brown and Levinson (1987) viewed politeness as an action that relates to the concept of face and considered factors such as power, distance, and imposition. Fraser (1990) looked at other approaches in which social norms and the obligations arising in conversation play a role. Ide (1989) highlighted the aspect of discernment: a particular form may be used since it is socially expected given the relationship, not because the speaker is voluntarily showing admiration. Subsequent work focusing on relationships and on discourse went even further. Watts (2003), Locher and Watts (2005), Haugh (2013), and Kádár and Haugh (2013) all oppose the idea that a surface form can be given a single interpersonal value without taking into account the participants, their histories, and the situations in question. A valuable result of this is methodological in that respect should not be deduced from po when the study is able to ask the speaker what function the form had.

Philippine studies make the caution less abstract since requests, refusals, letters written within institutions, and interviews given by broadcast media show how politeness decisions vary according to role, status, the nature of the task, and the medium (Agnir-Paraan, 2019; Boonkongsaen, 2013; Castro, 2018; AlAfnan & dela Cruz-Rudio, 2023; dela Cruz-Rudio, 2025). Santos (2022b) is particularly useful as it demonstrates that polite markers and honorific forms can be used in interactions which are also direct, adversarial, or face-threatening. The fact that they can coexist is important; a respectful form does not have to be incompatible with a difficult action.

Conversational AI provides a clear example to compare, yet it only offers a comparison. It is well known that people tend to apply social rules and show politeness towards computers (Nass, Steuer, & Tauber, 1994; Nass, Moon, & Carney, 1999; Nass & Moon, 2000). Generative AI is now able to produce interactions in which users can be either highly polite, highly efficient, or somewhere in between (Elsweiler, Ziegner, & Elsweiler, 2026). Large language models are also capable of mimicking a great many pragmatic features, even though they remain imperfect in terms of register, banter, context, and impoliteness (Andersson & McIntyre, 2025; Tantucci & Culpeper, 2026). A machine is therefore able to output Opo, naiintindihan ko po in a way that is perfectly recognizable. The form can be judged. The fact that this is so does not mean that the system has a human attitude of respect. This separation is what makes the AI example relevant to the current question.

The study investigates what occurs when the same separation is applied to human speech. There are four cases that are relevant: po is present and respect is reported; po is present and respect is denied; po is absent and respect is reported; and both po and respect are absent. The more difficult situations are the two in the middle. In the case where po remains present even though personal respect is denied, the issue is not whether the participant is inconsistent. Rather, the question is what function the po is serving. When respect is reported without po, the inquiry then becomes what other factors take on that stance. Factors such as age, office, habit, upbringing, restraint, distance, affection, irony, or the speaker's own rule about how to speak may eventually come into play. At this stage, however, they are still merely possibilities to be tested, not conclusions to be announced.

2. The Kind of Questions the Literature Enables the Study to Raise

2.1. The Form Is Part of a Social System and Not Merely Definable by a Single Word

Schachter and Otanes (1972) regard po and ho as enclitic particles which have both a grammatical position and a social function. Kess (1982) includes po, opo, and the other similar forms within a more extensive respect system. The idea is simple enough to incorporate in teaching: po stands for respect. This might be a useful way of explaining it to beginners. However, it is too simplistic for analytical purposes. A form that varies according to kinship relationships, age differences, social offices, the degree of familiarity, and the type of speech act cannot be properly understood merely by relying on a literal definition.

The analysis carried out by Gocheco in 2009 also reveals the same issue at discourse level. The particle po can be used to address an audience, to soften a statement, or to accompany an argumentative move. The particle is still identifiable even though the action associated with it changes. Sales (2016) includes a historical warning from missionary linguistics in that descriptions of politeness cues themselves are influenced by the categories that speakers and translators use when trying to explain them. A common interpretation may be correct without being complete.

Jeon (2026) offers a different kind of evidence. In a parallel corpus consisting of Korean and Tagalog, the honorific information in Korean did not result in a uniform use of po in Tagalog, the way the realization varied depending on the type of honorific material present. Although this result should not be used as a basis for a theory of Tagalog respect, it does make more difficult to justify the view that po simply reflects a fixed internal quantity known as respect; rather, the form is chosen within a linguistic and relational system.

2.2. To Say That Someone Is Polite Is Not the Same as Saying That One Has Respect for Them

The difference between using polite forms and showing personal esteem is found in various traditions within the study of politeness. Brown and Levinson (1987) concentrate on the management of face. Fraser (1990) demonstrates that politeness has also been interpreted in terms of social norms and conversational agreements. Ide (1989) highlights the idea of socially expected discrimination. Although these approaches disagree on some key theoretical issues, none of them holds that every polite form has to be seen as a direct expression of private admiration.

Relational work helps to define the boundary more clearly. Watts (2003) and Locher and Watts (2005) contend that conventional forms should not be regarded as polite by default when used outside the interaction in which they are judged. Kádár and Haugh (2013) see politeness as something that is part of relational histories and social practice, while Haugh (2013) stresses the importance of participation orders and the way in which evaluation is distributed. The significance for the current design is practical rather than far-reaching. The analysis must note what the participant said, what relationship was involved, what stance was reported, and what function was assigned; one column must not be used to replace the others.

That is also the reason why the study refers to 'participant-reported respect'. It does not assert that it has access to a private psychological state. An interview can note whether a participant says that a person is respected, not respected, or respected in one way but not in another way. It can record the reasons for using or omitting po. It cannot go beyond the report to directly measure respect. It is important to keep this limitation in view in order to avoid a circular result.

2.3. Difficult Acts Are More Informative Than Easy Ones

Making requests between people who don't know each other already requires a certain amount of social skill, but it doesn't place much strain on the traditional link between the use of 'po' and expressions of deference. It is disagreement, correction, refusal, and anger that do. Boonkongsaen (2013) discovered that status influenced the ways in which Filipino and Thai speakers used English to refuse. Both Castro (2018) and AlAfnan and dela Cruz-Rudio (2023) demonstrate how requests are constructed through self-presentation, mitigation, and supportive actions. Agnir-Paraan (2019) found similar relational choices in Philippine workplace memoranda. However, these studies do not provide evidence regarding every instance of the use of 'po'. What they do show is that communicative acts and institutional relationships affect how politeness is expressed.

The case of broadcast interviews is even clearer. Santos (2022b) records instances in which honorific and polite forms are used alongside confrontational ones. This combination is advantageous since it avoids a false dilemma. An utterance need not be either deferential or challenging; it can be both in different ways. Culpeper (2011) makes the same methodological observation in the field of impoliteness research: context, conventionalization, intention, and uptake are all important. Sarcasm cannot be reliably identified on the basis of a single word and anger cannot be deduced simply because a particular marker is missing. The protocol therefore requires participants to explain the connection between the wording and the stance rather than having the codebook determine this relationship in advance.

Santos (2022a) also warns against considering the imported models of politeness as complete descriptions of interaction among the Filipino people. The fact that this warning is given does not serve to establish a theory that is uniquely Filipino merely by stating so; instead, it demands that evidence be obtained from speakers of Filipino and from Filipino contexts. The present study adopts a modest approach by asking adult speakers of Tagalog how they use and understand these forms in controlled contrasts.

2.4. Elicitation Can Test Contrasts, but Not Frequency in Ordinary Life

Controlled scenarios are useful since the relationship can be altered while the task stays roughly the same. This has always been one of the reasons for the use of discourse-completion tasks and other elicitation methods in the study of pragmatics (Blum-Kulka, House, & Kasper, 1989; Kasper & Dahl, 1991). Félix-Brasdefer (2010) demonstrates that different elicitation and verbal-report methods can be combined provided that their evidential functions are kept separate.

The limitation is not a footnote. Golato (2003) shows that elicited responses and naturally occurring interaction do not answer the same question. A participant who writes Mali po kayo in a scenario has shown what that participant produced under that prompt. The response is not a frequency estimate and it is not proof of what would happen in an unobserved real dispute. The follow-up interview adds explanation, not hidden access to the mind. The protocol is built around those limits rather than trying to write past them.

Semi-structured interviewing is used for the same reason. Kallio, Pietilä, Johnson, and Kangasniemi (2016) emphasize guide development, sequencing, and pilot testing. Here the participant’s own first response becomes the object of the later question. If po appeared, the interviewer asks why it belonged there. If it did not, the interviewer asks why it was unnecessary. This makes the interview answer a different question from the initial production task.

2.5. Digital Talk and AI Cannot be Collapsed into One Variable

A prompt that is typed is not just spoken language entered on a keyboard. As Walther showed in 1996, communication through computers can affect the way cues are given, messages are constructed, and interpersonal interaction is presented. Herring (2004) makes a similar argument, claiming that it is necessary to analyze specific online discourse practices rather than making generalizations about a single digital register. This is also true in the Philippines: in the area of higher education, written requests exhibit their own distinct patterns of politeness (AlAfnan & dela Cruz-Rudio, 2023; dela Cruz-Rudio, 2025).

This presents a problem in the comparison with AI. When a participant omits the word 'po' while typing into an AI system, the omission could be due to the machine itself, or it might be due to the act of typing, the participant's focus on the task, or simply to their usual style when making digital requests. That is why a brief check is made regarding spoken versus typed input. It will not allow for the conclusion of a general medium effect. It will only serve to avoid the paper attributing every difference to the AI.

Earlier literature on human-computer interaction has demonstrated that it is possible for people to be polite to machines without considering them to be persons in the full human sense (Nass et al., 1994; Nass et al., 1999; Nass & Moon, 2000). More recent studies have built on this idea in the case of generative AI: there are considerable differences in users' politeness (Elsweiler et al., 2026), while the way the models deal with impoliteness and pragmatic cues is inconsistent and depends on the context (Andersson & McIntyre, 2025; Tantucci & Culpeper, 2026). The current paper does not need to make any more claims since AI is useful in that it distinguishes recognizable linguistic politeness from any automatic assumption of human respect on the part of the system.

2.6. The Remaining Question

There is already sufficient reason to oppose a simple method of counting and then drawing a conclusion. Tagalog has conventional forms of respect and the use of these is socially structured. The degree of politeness differs according to the relationships involved, the actions taken, and the medium. The way in which participants evaluate each other is important. It is also possible for people to direct their social and polite behaviour towards machines. The aspect that has not been examined closely enough is how Tagalog-speaking adults themselves explain the cases involving a mismatch: instances in which po is retained even when no personal respect is reported, cases where respect is reported without po being used, and cases in which po is carried over into interaction with conversational AI.

The proposition under consideration is intentionally kept simple. The use of po or opo shows that the conventional form was employed; however, on its own it does not prove that participants reported respect. The lack of the form shows only that it is absent; by itself it does not indicate disrespect. The empirical work starts only after these two cautions, not before.

3. Research Questions

The main research question is:

How do Tagalog-speaking adults distinguish the use of po and opo from participant-reported respect and other relational functions across human and AI-mediated communication?

• At what times are po and opo kept in cases of disagreement, correction, irritation, obligation, distance, or sarcasm?
• At what times do participants state that they respect the subject even though neither po nor opo is present?
• Why do the participants give as reasons for using po and opo towards people whom they state that they do not personally respect?
• What influence do age, authority, familiarity, conflict, and the nature of the communicative task have on the use or interpretation of po and opo?
• In their typed interactions with conversational AI, how do participants make use of and interpret po and opo?
• When participants direct po at the AI, do they describe this practice as being aimed at the machine, the interactive situation, a habit or their upbringing, their own way of speaking, or something else?
• Is the relationship between form and function different when the task is carried out orally rather than by typing?

4. Method

4.1. Design and Evidential Boundary

The approach involves four linked tasks: scenario elicitation, a stimulated semi-structured interview, paired utterance judgments, and a small spoken-versus-typed comparison. The tasks are linked since none of them on their own can deal with the entire question. Scenario elicitation sets up controlled contrasts between relationships. The interview requires the participant to account for a real response which has just been given. In the paired items the wording is kept almost the same but the honorific marking or the stated stance is changed. The mode check serves to ensure that effects due to typing are kept distinct from those due to the person being addressed. Research into pragmatics has supported the use of elicited comparisons while at the same time cautions that elicited speech is not the same as speech that occurs in natural situations (Blum-Kulka et al., 1989; Kasper & Dahl, 1991; Félix-Brasdefer, 2010; Golato, 2003).

The study does not initially specify po and opo as its narrow focus; instead, participants are informed that it is about communication in different situations. If the terms had been named before the first response was made, the learned definition would have become particularly clear and might have caused the participants to give the expected answer. The focus is made clear only after the questions concerning po and opo themselves have started. This order of presentation conforms with the general requirement that interview procedures should be based on the evidence one is trying to obtain and that they should have been tested in pilot work (Kallio et al., 2016).

The first response is not regarded as more truthful than the later explanation; they are merely different kinds of records. The one illustrates what was produced in response to a prompt, while the other shows how the participant accounts for that production after having given it attention. Neither of them provides a direct measure of an inner state.

4.2. Participants and Sampling

The participants must be adults who regularly use Tagalog or Filipino in their day-to-day interactions. The background information to be collected will cover their age, their primary and home languages, where they were brought up, how often they currently use Tagalog or Filipino, their occupation or whether they are students, and how frequently they use conversational AI. Even if Tagalog or Filipino is part of their usual range of languages, multilingual speakers from the Philippines are not excluded and their language background will be taken into account in the analysis. It will not be assumed that findings from Tagalog speakers can be applied to the Philippine languages as a whole.

For the pilot study four to six adults will be recruited. Recruitment will then take place in batches, the first of which is expected to involve about twenty-four to thirty interviews. This range is not considered to be a fixed figure. The question of whether a qualitative sample is adequate depends on the research aim, the degree of sample specificity, the quality of the dialogue, and the level of analytic development that is desired (Malterud, Siersma, & Guassora, 2016). Research into saturation also makes a distinction between the early emergence of recurring codes and their later development of meaning (Guest, Bunce, & Johnson, 2006; Hennink, Kaiser, & Marconi, 2017; Hennink & Kaiser, 2022).

The collection process will cease when carrying out further interviews no longer has a material effect on the description of the main cases involving a mismatch and when there is sufficient evidence to allow the principal relational contrasts to be compared. A new participant who simply reproduces an existing label contributes less than one who demonstrates that the same label is interpreted differently depending on age, authority, closeness, or conflict. The decision to stop will be recorded rather than being deduced solely from a predetermined number.

4.3. Scenario Elicitation

The scenarios differ in terms of relationship and address while still keeping the communicative task as similar as is practical. They involve a parent, an older stranger, a younger authority figure, a close friend, an older person that the participant does not personally respect, a formal service encounter, and a conversation with a conversational AI system. Requests, disagreements, corrections, and responses to errors are all included since the expression of politeness depends on power, distance, status, and the nature of the communicative act (Brown & Levinson, 1987; Blum-Kulka et al., 1989; Boonkongsaen, 2013).

The instruction is meant to be simple: write something that is most likely to actually be said or typed, not attempt to give the correct answer. In no situation should po, opo, kayo, a kinship term, sir, ma’am, or any other form of politeness be used. The responses are kept exactly as they were before coding. What is needed is for participants to make spontaneous choices in response to a controlled prompt, not evidence that they know the textbook explanation of po.

4.4. Stimulated Semi-Structured Interview

Following the initial task, the participant's own words form the basis of the interview. When either 'po' or 'opo' is given, the interviewer asks the reason why it had seemed natural in that response; if neither 'po' nor 'opo' is present, then the interviewer inquiries as to why it had not been necessary. The follow-up questions then modify one factor at a time—age, authority, closeness, anger, disagreement, the absence of personal respect, or the type of addressee. Although the main guide stays unchanged, the sequence of the probes can be in line with the participant's answer, in accordance with the practice of semi-structured interviews (Kallio et al., 2016).

One branch is more important than the rest. A participant might essentially state that a person is not personally respected yet is still addressed with 'po'. The response is not treated as a contradiction. The interviewer then asks what the person is doing in that situation. If the participant gives as an explanation office, age, upbringing, restraint, habit, distance, or a rule concerning personal conduct, that explanation is noted. If the participant offers a different reason, the coding framework has to allow for it. Candidate functions do not turn into results just because they were expected.

4.5. Paired Utterance Judgments

The participants are to compare a small number of pairs of utterances. It is not necessary to decide which version is more polite and which one is less polite; rather, the aim is to examine what happens when an honorific form is added, taken away, or placed into a different kind of relational context. By means of elicited comparison such contrasts can be identified, on the condition that the judgments are not taken to reflect natural conversational frequency (Kasper & Dahl, 1991; Golato, 2003).

Hindi po ako sang-ayon. / Hindi ako sang-ayon. - “I do not agree,” with and without po.

Mali po kayo. / Mali kayo. - “You are wrong,” with and without po.

Salamat po. / Salamat. - “Thank you,” with and without po.

Ingat po kayo. / Ingat, Nay. - Two relationally different ways of saying “Take care.”

Opo, kayo na po ang tama. / Sige, kayo na ang tama. - A respectful-form version and a non-honorific version of “Fine/Yes, you are right.”

In each case, the participants are asked to state the changes in tone, distance, role, or relationship, and also to consider whether one or other of the versions could be seen as respectful or disrespectful depending on the situation. The phrase Opo, kayo na po ang tama is given twice with the same wording in both instances: one instance reflecting sincere acceptance and the other showing irritated or sarcastic surrender. A change in judgment when the context is altered would explain why the wording alone is not sufficient to determine the interpretation. According to relational approaches to politeness, this kind of test is more defensible than giving a priori politeness ratings (Locher & Watts, 2005; Haugh, 2013; Kádár & Haugh, 2013).

4.6. Spoken-Versus-Typed Comparison

In a few of the matched situations the response will be given both orally and by means of a typed message. The comparison is diagnostic and not based on the entire population. Computer-mediated interaction can influence planning, the availability of cues, and interpersonal presentation (Walther, 1996; Herring, 2004), and research into email in the Philippines indicates that written requests have their own patterned politeness practices (AlAfnan & dela Cruz-Rudio, 2023; dela Cruz-Rudio, 2025).

An explanation based on AI becomes less plausible when a participant fails to include 'po' both in interactions with AI and also in similarly typed responses directed at a human. On the other hand, if 'po' is present in responses directed at a human but is absent when interacting with AI, the difference in addressee becomes more significant. Hence, mode and addressee are treated as separate categories.

4.7. Human-AI Comparison

The participants are asked by the AI to ask for help, to correct an incorrect answer, and to react to an AI response which includes po and opo. The study does not investigate whether the machine actually respects people. This question is unnecessary since previous research has already demonstrated that people are able to respond to machines in a social and polite manner (Nass et al., 1994; Nass et al., 1999; Nass & Moon, 2000). More recent research into generative AI shows that there is a great deal of variation in users' politeness and that the models handle pragmatic context in an uneven way (Elsweiler et al., 2026; Andersson & McIntyre, 2025; Tantucci & Culpeper, 2026).

The appropriate evidence is the participant's explanation. Someone who types 'po' into an AI might claim that the form is intended for the machine, for the conversational context, for habit, for the way they have been brought up, or for their own personal standard of speech. Another participant, on the other hand, might reject all of those explanations. The comparison is useful exactly because it makes the interpretation open to debate rather than obvious.

4.8. Analysis

From the beginning, coding separates the four dimensions: FORM, RELATION, REPORTED STANCE, and ATTRIBUTED FUNCTION. The form po is coded as po; it is not considered to be coded as respect just because it occurs. Absence is coded as absence, not as a lack of respect. Both respect and non-respect depend on a report by a participant. Sarcasm requires either an attribution by a participant or an explicitly stated sarcastic condition. Although these rules are simple, they stop the focal form from serving as its own evidence.

Cases involving a mismatch receive the most careful examination. If po is present but personal respect is refused, the analysis looks into the reason for retaining po. In the case where respect is reported but po is missing, the analysis investigates what other relationship, formulation, or practice is said to convey that respect. Age, role, habit, upbringing, distance, restraint, affection, and speaker self-conduct are all possible ways of coding it provisionally; these are only kept when the data provide repeated support for them.

The analysis is carried out in the following order: by utterance first, then by explanation second and comparison third. When coding the response it is not necessary to include the later interview explanation. The explanation is coded on its own. The cases are compared in those instances where the wording is the same but the relationship changes, in those where the relationship is the same but the stance changes, and in those where the stance stays the same but the honorific marking changes. This ensures that the elicited performance is kept separate from the retrospective account (Félix-Brasdefer, 2010; Golato, 2003).

The audit notes should include records of code changes, counterexamples, and cases which do not fit the developing account. It is not necessary for the paper to have a single function. A defensible outcome could be that the same form carries out different functions in different relations. The existing relational accounts of politeness already make this possibility entirely plausible (Locher & Watts, 2005; Haugh, 2013).

4.9. Pilot and Decision Rules

The pilot phase is not merely a brief summary of the results but a test of the instrument itself; it fails when participants require a great deal of explanation in order to understand the situations described, when the intended social contrasts cannot be clearly identified, or when the direct questions elicit only statements that merely convey respect. Pilot testing forms part of the careful development of the interview guide since difficulties with the wording or the order of the questions are easier to correct before the main data collection takes place (Kallio et al., 2016).

The items will also be revised in such a way that they incorporate the desired response directly into the prompt. It would be impossible for a scenario to state that the participant still respects the addressee and then use that scenario to find out whether respect is in fact present. An AI item must not refer to the system as being respectful before asking how its po is interpreted. The instrument should provide the circumstances needed for a judgment to be made, not include the judgment itself.

Three boundaries are fixed: the study will not draw a conclusion about a national trend from a qualitative sample; it will not infer a psychological state from po alone; and it will not say that AI causes a decline or change in the use of Tagalog honorifics on the basis of data from cross-sectional interviews. Making such claims would require different evidence.

4.10. Ethics and Participant Protection

Taking part is optional; questions can be left aside and involvement can cease without any penalty. Names will not appear in the analytical files and quotations which are intended for publication will be de-identified. Permission must be given explicitly if audio recording is to take place. Since certain scenarios involve a person that the participant does not respect, it is not necessary for anyone to give that person's name and a general reference can be used or the item can be refused.

During the first elicitation task the only aspect that is not focused on is po and opo, and this is done merely to reduce priming. The fact that this aspect is not focused on is made clear before direct questioning starts. The risks, the recording, the use of the data, the voluntariness, and the general purpose of the research are not hidden.

5. What the Study Can Establish

The paper starts with a minor claim and concludes on a similar scale. The use of a conventional honorific shows that the form was in fact employed. It does not provide enough evidence to prove that respect was reported by the participants. The lack of that form is also not sufficient to prove that there was no respect. Tagalog descriptions show that po and opo are part of respectful and deferential address (Schachter & Otanes, 1972; Kess, 1982; Gocheco, 2009). Research into relational politeness explains why the assessment of an utterance cannot be based solely on one marker (Watts, 2003; Locher & Watts, 2005; Haugh, 2013; Kádár & Haugh, 2013). Rather than deciding between those propositions, the study bases itself on the participants' data.

If people continue to use po towards those whom they have explicitly stated that they do not personally respect, the result would not prove that po is no longer a form of respect; it would only show that the conventional form can still be used while fulfilling a function other than the direct expression of personal esteem. The precise details would have to be provided by the participants themselves. If respect is reported in the absence of po, then the finding would indicate that the stance can also be conveyed through other relational or linguistic resources.

The use of an AI comparison helps make the distinction more apparent. It is possible for a system to generate a respectful and well-formed response, and a user can then speak to that system in a polite manner, without having to decide whether or not human-like respect should exist within the machine. The explanations given by participants might indicate that people's polite attitude towards AI stems from habit, the way the conversation is framed, their own behavior, the machine itself, or some factor that is not covered here. The comparison has value only if the paper keeps these possibilities open.

The fact that these limitations are not flaws to be concealed is that the study is unable to measure private respect directly; it cannot estimate how frequently po and opo occur nationwide; it cannot demonstrate a historical decline based on a single cross-sectional sample; it cannot prove that AI has altered Tagalog; and it cannot convert the elicited scenarios into actual recordings of everyday life (Golato, 2003). The only thing the study can achieve is more limited in scope: it can show how Tagalog-speaking adults relate or separate honorific forms, social relationships, reported stance, and attributed function when those elements do not align neatly.