Reluctant heroes: New doctors negotiating their identities dialogically on social media

Ensuring that students transition smoothly into the identity of a doctor is a perpetual challenge for medical curricula. Developing professional identity, according to cultural‐historical activity theory, requires negotiation of dialectic tensions between individual agency and the structuring influence of institutions. We posed the research question: How do medical interns, other clinicians and institutions dialogically construct their interacting identities?

of doctors in training and propose that, to maintain the vitality of medical education, institutions should seek to reconcile their projected identities with the lived identities of graduates.

| INTRODUCTION
The transition from medical student to intern does not always proceed smoothly. 1In health care systems where interns make important contributions to hospital medical services, 2 rough transitions into practice can compromise both patient care 3 and interns' own well-being. 4Researchers who framed this problem in terms of interns' preparedness for practice found that up to one-third of interns are inadequately prepared. 5,6][6][7][8][9][10][11][12] Feeling that 'they had not been trained adequately in management aspects of the job' tested interns' 'fortitude to work'. 1 More clearly understanding tensions that influence interns' entry to practice could help medical education have a stronger impact on health care.
Engeström's cultural-historical exploration of expertise in transition notes that clinical novices 'often make choices based on fulfilling their identities and following rules and routines that they themselves may not be aware of'. 13Monrouxe argued that it is helpful to regard identities in medical education not as fixed cognitive schemas but as the product of complex interactions where language and patterns of behaviour mediate the identity formation of both individuals and institutions. 14Medical education researchers have used a variety of method(ologie)s to respond to her call for further examination of how 'identities are constructed, enacted, invoked, or exploited in a variety of interactional settings'.They have developed categories describing the experiences of medical students becoming residents 15 from no specific theoretical position.Viewing identity from the perspective of values and emotions, they have reported adverse effects of a hidden curriculum. 16A psychological perspective guided researchers to frame identity as an assessable competency to be inculcated pedagogically, 17,18 an attribute that complements but is distinct from competence, 19 and, from a social psychology perspective, a product of the structuring influences of group dynamics. 20,21Researchers have treated professional identity formation as a phenomenon that influences physicians' employability 22 and used narrative methodologies to explore students' experiences of entering clerkships 23 and residents' experiences of transitioning into specialty training. 24This body of work, even when acknowledging social influences, has tended to represent identity as an individual attribute rather than a complex, dynamic process that co-involves individual and social milieu.
Cultural-historical activity theory (CHAT), explained in the Conceptual Orientation section of Methods, locates identity within a bidirectional flux between individuals and social milieus.Research oriented towards Foucault's critical theory 25,26 assumes that culture and history form relatively stable 'social discourses' that shape the identity possibilities available to individuals.An alternative critical discourse methodology, dialogic analysis, conceptualises identity formation as a process in which individual identity and social discourses of identity are in dialectic tension.This perspective affords greater agency than a Foucauldian perspective for individuals to influence their own identity development.Researchers in the dialogic tradition have shown how students engage in 'creative struggle' and resist subjugation as they develop relational identities. 27They have shown how supervisors' feedback on residents' behaviour constructs competences (notably collaboration and communication) in ways that reinforce doctors' high status in relation to patients and members of other clinical professions. 283][34] They have examined how clinical supervisors create identity possibilities for students that widen or narrow the hierarchical distance between doctors and patients. 35,36]39 Recently, social media have made speech acts available to researchers in new ways.Microblogging tools such as Twitter provide an open-source social commentary, within which dialectic tensions that influence identity formation are dynamically enacted.Empirical evidence suggests that microblogs can, indeed, be an informative resource for qualitative researchers.A large archive of Twitter posts made newly qualified doctors' informal professional socialisation accessible to researchers. 40Twitter provided a resource for researchers to explore the complex relationship between risk and reward in expressions of gratitude towards health workers during the COVID pandemic using discursive psychology. 41Microblogging has also been studied dialogically.One study used corpus linguistics to examine speech genres in Twitter comments on the Panama papers 42 ; another compared freelance journalists' and physicists' use of posts to communicate knowledge during the 2011 Fukushima nuclear crisis in Japan. 43 reasoned that the COVID pandemic was an 'experiment of nature' that, by making preparedness a pressing social issue, could provide transferable insights into identity development when students become interns.Dialogic analysis of microblog feeds could help us explore dialectic tensions.We aimed to provide curriculum leaders and regulators with information that could, ultimately, help smooth interns' transitions into practice, to the benefit of them and patients.
The research question was: How do interns, other clinicians and institutions dialogically construct their interacting identities?ORIENTATION

| Ethical approval
The Research Ethics Committee of the Faculty of Medicine, Health, and Life Sciences, Queen's University Belfast, approved the project (Reference MHLS 20_64).Noting that research committees give inconsistent opinions about researching social media, commentators have recently proposed that researchers should exercise due diligence in conducting their research, paying attention to the balance between potential benefit and harm with special attention to privacy and consent. 44Our research ethics approval did not require us to obtain consent to include Twitter posts in our research or quote them in this article because posts are public statements.Wishing, however, to minimise the risk of deductive disclosure of individuals' identities, we chose 10 posts that were particularly relevant to the research question to quote verbatim and followed the accounts from which these had been posted.Five of the 10 account holders followed us back (the other five did not reply), which enabled us to send direct messages asking permission to quote their posts verbatim.All agreed and only their posts are quoted in their entirety.We included the content of 177 other posts in our analysis.To disguise the identities of these account holders, we have woven short phrases or words into a narrative presentation of results tagged with a unique code linked to a master spreadsheet, which is available for audit.

| Theoretical orientation
A recent overview of CHAT from its origins in Marx and Hegel through to its operationalisation in third-generation activity theory 45 provided content for the following explanation. 46We use bold font to identify technical terms specific to CHAT, which is a philosophy of human existence and a theory of learning.It regards existence and consciousness as located within activities, which are essentially social and communal, and imbued with the cultural history of the communities in which these take place.Activities are made up of actions, which may contribute to more than one activity.Actions are mediated by material artefacts and symbols, themselves products of cultural history.Principal amongst these symbols is speech, spoken or written.Speech mediates consciousness and thought.From the perspective of CHAT, actions are, of their essence, imbued with purpose, whose motivating force is the object of an activity.The object may be regarded as the raw material, or problem space towards which an activity is directed.
Bakhtin, a literary critic and near-contemporary of Vygotsky, further developed a cultural-historical conceptualisation of how speech mediates social action. 47According to his theory (dialogism), humans conduct their lives dialogically: Speech acts purposefully address other people, even absent or imagined ones, and respond directly or indirectly to earlier addresses.This concept is termed addressivity.
Bakhtin theorised that utterances (continuous speech acts from when a person starts uttering to when they finish) are complete 'units of analysis': purposeful, completed, acts of agency, which give researchers valid insights into human development.Much of the dialogic research in medical education, to date, has been oriented towards Holland and colleagues' cultural-historical Figured Worlds theory 48 of identity development.Figures, people we encounter, present possibilities for our own identity, which is in a constant state of flux.Speech acts allow us to author identities, though our freedom to do so is constrained by positions made available to us by institutions and other people.Using these assumptions to guide the analysis of speech acts is referred to as critical discourse analysis, where the term 'discourse analysis' refers to the analysis of dialogue and the term 'critical' refers to the dialectic assumption that power and goods are unevenly distributed in society, making social possibilities unequally available to people.Gee has provided a set of linguistic tools, described in detail elsewhere, 49,50 that operationalise the analysis of utterances.He did not intend these to be applied slavishly; rather, they sensitise researchers to use of grammar, metaphor, implied meaning and other linguistic features.Sullivan has offered epistemological guidance to dialogic researchers wishing to craft research findings from dialogue. 51Although a minority discourse methodology in health professions education research, dialogic discourse analysis is an emerging theme (see, e.g. 30,32,35), as is the analysis of social media posts.Together, this methodology and data source form a consonant whole: Posts are utterances, imbued with the purpose of a person addressing other people and inviting a response.Strands of posts are dialogues, making the flux of human identity construction and the enablement or constraint of it by individuals and institutions accessible to researchers.The flux of dialogues shows agency and power at play in ways that make social positions and identities more or less available to people in particular contexts and moments of time.

| Setting
Although we screened for Twitter posts from all over the world, actively seeking ones from other countries or cultures, almost all relevant ones came from the UK.Only a small number came from Ireland and the USA.

| Research team
TD is a hospital specialist physician and education researcher with special expertise in CHAT research.He was amongst the first medical education researchers to use Figured Worlds and developed the dialogic methodology used here.DA was a medical student who did the fieldwork during a 1-year BSc intercalation between the third and fourth years of the Queen's University Belfast medical programme.DB, whose clinical speciality is general practice, leads an undergraduate programme in Ireland and is an education researcher with special expertise in dialogic research.HG was a trainee education researcher and intern at the time of the research.HR is a clinical senior lecturer in general practice and medical education researcher with expertise in critical qualitative research including dialogic analysis.

| Study design and timescale
The research began in March 2020 when it became obvious that COVID would accelerate the transition of some students to interns.
Table 1 shows the design and timescale of the project.Data collection began when the 'accelerated-interns' (A-interns) entered practice.
Twitter posts dwindled to negligible numbers within 2 months.Continued monitoring confirmed that the dialogue had stopped until 'conventional track interns' (C-interns) entered practice and ended, again, within 2 months.

| Screening and selecting Twitter posts
We chose to identify relevant posts by monitoring rather than requesting a download of Twitter feeds for many reasons: data were readily available; the sheer size of a download could swamp detailed analysis; the interpretive nature of the research did not require representative sampling; monitoring posts and reflexively choosing informative ones would produce a rich and manageable sample; and a monitoring/reflexive sampling strategy would make it possible to follow conversations longitudinally.DA set up a study Twitter account in her own name and used the Advanced Search function to identify posts.She used the research question and inclusion criteria (below) to define relevance, discussing with other team members when in doubt.
In addition to searching, she purposively browsed posts from universities, training bodies, regulators and other interested parties in Englishspeaking countries around the world and the Nordic countries, Germany and the Netherlands where English language fluency is common.As well as following all promising sources of posts, she identified relevant hashtags (such as #tipsfornewdocs).

| Inclusion and exclusion criteria
The inclusion criteria included the following: • open unlocked Twitter account; • topic related to final-year medical students, new doctors and preparation for entry to practice; and • accounts of individuals or organisations (lay and professional), within or outside the health care education sector.
Irrelevance was the only exclusion criterion.

| First screening
DA forwarded each post to all team members using Twitter's private message function.We debated suitability by messaging, e-mail and holding regular online meetings.

| Second screening
Seven days after a Tweet was first posted (to allow responses to be posted), DA distributed posts equally between team members, who reviewed eligibility for inclusion, completed the form described below and did a first-pass dialogic analysis.

| Data entry and storage
A Microsoft form allowed the team to store the texts of posts, code them, review posts entered and coded by others and overview the dataset.The form included, for each post, the following information: • the account holder's handle; • its text, including visual elements: emojis, GIFs and other pictures; • the point on the timeline shown in Table 1; • the identity of the user and their location; and • the team member's analysis of significant linguistic features; languages and discourses represented by the text; the significance of the post within the thread of which it was part; and up to five related posts that followed it.

| Data handling for analysis
We aggregated data in an Excel spreadsheet, to which extra columns were added to record interpretations and comments by other team members.

| Methodology: Discourse analysis
We used a dialogic qualitative analysis methodology as theorised above, 51 developed by us, 33,35,36 that has been applied successfully to similar, fragmentary data. 33This depended on close, repeated reading of textual data by more than one person, writing memos offering tentative interpretations and frequent in-depth conversation between researchers.Given the constructionist theoretical orientation of the work, it depended, also, on researchers' reflexivity.This means being curious about why language was being used in one way and not another way and asking ourselves what identities and social possibilities are constructed by speech acts. 49,50In this context, it also meant using our familiarity with clinical practice, education, work contexts and dialogic theory to understand language and consciously trying to 'make the familiar strange'.We agreed that the sample was sufficient because extensive and repeated whole-team discussions had led us to identify a clear, consistent, dialogic structure in the first 2-month period, which posts in the second period replicated.

| Study procedures
See Table 2.

| Reporting the findings
The spreadsheet contained 182 Tweets, 82 of which were solitary and 100 of which were in dialogic chains.We termed the first post in a chain (or a solitary one) the 'Initial' one (acknowledging that addressivity theory contests that any utterance can ever be truly initial) and subsequent ones 'responses'.Our use of the word 'identity' is guided by the assumption that making any utterance does identity work. 53en it was clear that a person had posted a comment on behalf of an institution or was addressing one, we used the term 'institutional identity', otherwise assuming that a post constructed the identity of an individual, albeit in a social context.We identified these individuals as senior doctors (fully trained), residents (most of whom were junior This quotation exemplifies language early in the pandemic when medical students' graduation was accelerated.They were 'heroes' 6ININST, 10ININST , 'the future of medicine' 4ININST, 9ININST , 'our greatest achievement' 6ININST and 'jewels in the crown of the health service' 9ININST , who would 'do us proud' 12ININST by being 'valuable members of teams' 12ININST .This sentiment was echoed by alumni who said graduates of their medical school would 'play a role in shaping the response to the pandemic' 55ININST .Institutional posts contained strongly affective symbols and language.
Emojis-clapping hands 44ININST and a prayer sign 13ININST -were coupled with expressions of respect and gratitude towards new interns.Institutions congratulated new doctors, were 'proud' of them 44REINST, 55ININST , 'loved seeing the smiles on their faces' 44REINST , welcomed them to the profession 44ININST and 'trusted them' 12ININST .The debt of gratitude 'owed by the nation … could never be fully expressed' 13ININST .Institutional posts attributed agency to interns by saying they had 'volunteered to start their medical careers early' 10ININST, 44REINST and 'moved into practice' 6ININST to 'care for patients' 44ININST .Posts did not specify how institutions had contributed to interns' agency.A single post from a health care provider hoped their induction course would help interns 'settle in well' and they would be 'ready to ask many questions' 111ININST .T A B L E 2 Study procedures.

Using methodological tools
Team members read, presented at team meetings, discussed and applied concepts of dialogism described above (Theoretical Orientation) and familiarised themselves with several of Gee's heuristics to identify: • social languages and speech genres; • situated meanings; • Figured Worlds; • building of activities and identities; and • agency ('doing and not just saying').

Analysing utterances
We used a process of constant comparison to ensure that our evolving interpretations remained true to the raw data, using the linguistic heuristics to justify interpretations.We identified: • the social position from which the person posting is making their utterance; • how the text constructs their positional identity; • who they might be answering and addressing, paying attention to wider discourses; and • in what ways a post might be hegemonic or exemplify struggle.As the goal was to move beyond rigorous analysis to synthesising an interpretation, we regard interim frameworks as 'throwaway prototypes', which contributed to but do not constitute reportable findings (these are available by request to authors, but not in a repository to protect account holders from deductive disclosure).

Team working
We met every 1-2 weeks during April-May 2020 and regularly during the intervening and following periods, then met and/or corresponded to discuss drafts of this article.The main purpose of these meetings was to support a seamless process of individual and team reflexivity, 52 which consisted of constantly comparing interpretations with original posts, reflecting on, questioning each other about and discussing interpretations before agreeing and recording these in the data spreadsheet.This process progressed from analysing individual utterances to identifying similarities and differences between them and synthesising an increasingly comprehensive, agreed interpretation.

Synthesising an interpretation
Trial analysis: At the study midpoint, we all read the entire dataset and identified commonalities in the texts.This produced a first-pass interpretive framework.First comparative analysis: Next, we analysed data using the framework and discussed the findings and differences in analyses as a team.This produced a second-pass framework.Second comparative analysis: We repeated the coding, using the revised framework and adding comments to explain their interpretations.Revision of the framework: To facilitate review and revision, we represented the framework as a mind map, reviewed this against the raw data and adjusted its structure.Third comparative analysis: We divided the dataset between team members, each analysing part of the dataset jointly with another team member, identifying scope to merge, split, remove or add elements of the framework, which we reviewed as a team.Developing descriptive rubrics: We developed a tentative set of rubrics to describe all the elements of the framework, to help ourselves do a further constant comparative check.Fourth comparative analysis: We now paired with different team members to critique a further condensed version of the framework and identify posts that epitomised the findings.
Report writing TD synthesised a final interpretation of the framework's contents, drafted this narrative report of findings and devised Fig. 1, which co-authors critiqued before final revision.
Interns' posts, by contrast, specified how medical schools had not helped them develop agency by not 'preparing you for internship' 81ININTERN .Assessment of arcane knowledge had dominated medical school experiences 13ININTERN, 060ININTERN , leaving interns without practical knowledge, which experiences of 'shadowing' practitioners had not rectified 73REINTERN .Interns' posts used irony and negative affective language to construct a non-heroic identity for themselves.
Their lack of a legitimate doctor's identity was expressed as 'a totally reassuring new job title (side-eyes emoji)' 25ININTERN , a 'crushing realization' they would soon be a doctor 104ININTERN , and regarding it as progress that they had 'even managed to admit they were a doctor' 56ININTERN .The excitement of doing 'real medicine' was coupled with terror of 'not being able to complete simple tasks' 50REINTERN .They 'assumed we all don't know what we're doing' 32ININTERN , found they were 'as useful as a chocolate teapot' 50ININTERN and were only capable of 'sending the whole ward into utter chaos' 72ININTERN .They would be 'useless at answering the phone' 30ININTERN and 'scared of having to do so' 30ININTERN .They were capable only of 'smiling, asking for help', keeping job lists and not feeling bad about being stressed 43ININTERN .They were 'terrified of angering nurses'; successfully inserting a cannula was a 'miracle' 74ININTERN .
The identities constructed by these posts were freighted with negative emotions.As responsibility 'loomed' 64ININTERN interns had an 'impending sense of doom' 45ININTERN and were 'as nervous as hell' 45ININTERN .Starting internship was 'an impending inferno', not just a 'baptism of fire' 91ININTERN .Interns used emojis, other images and ironic language to construct their less-than-heroic identities: a toddler with a stethoscope 60ININTERN , a baby meerkat in a cot wearing pyjamas 73ININTERN , a clown wearing a party blower hat 18ININTERN , a nervous-looking Homer Simpson 70ININTERN , a loudly crying face 110ININTERN and a sad face streaming with tears like a child on their 'first day of primary school' 71ININTERN .As a student, 'watching a doctor make a decision was exciting' but, as an intern, 'making a decision Social support to interns' fragile identities elicited strong positive emotions.They were 'grateful for patience and kindness' 47ININTERN , described the clinical team of which they were part as 'SO helpful' 46ININTERN or 'bloody lovely' 30ININTERN , spoke of 'amazing staff who were incredibly supportive' 36ININTERN and 'blessed the wee cotton socks' of an ancillary worker who defended them against a hostile resident 79ININTERN .

| Residents and other team members
'Honestly, no one expects you to know anything.Sorry if that sounds brutal but everyone working in a hospital is expecting to do a lot of hand holding and helping for the next few weeks.People did it for us, we do it for the next lot.Tis the circle of life.You'll be reet (all right).' (Quoted with permission, resident 86RERES ) A resident's response to the intern 86ININTERN whose 'humungous imposter syndrome' is quoted above was that 'only 10% is stuff from medical school', the rest being, mainly, mastering the logistics of workplaces but also 'figuring out senior doctors' whims' and 'bluffing'

86RERES
. For another resident, only 5% of medical school was 'learning and were inspired by the willingness of interns to undertake their first night shift, which would be 'a service' 10INSE .Some posts explicitly positioned seniors high in the hierarchy of medicine by, for example, describing interns as 'young 'uns' with 'different risk-taking thresholds' 12INSE and agreeing that 'we need to look after them' 12RESE .

| Interns and residents
On call 'as a doctor' could be 'terrifying, fun, difficult, and stressful' 95ININTERN .Although 'people said it's a scary time', it could be 'an exciting and positive experience![Glowing star emoji]' 69ININTERN .
Posts described interns as being practically and emotionally 'OK' 65ININTERN .'Lots of senior support' and 'plenty of opportunities for impromptu teaching' helped them 'enjoy the first week as a doctor' 53ININTERN , which was 'SO much better than being a medical student' 68ININTERN .Nevertheless, interns 'lost nights ruminating over patients' 40ININTERN .Help with 'performing a no 1 nerve-inducing Dr task' helped an intern learn it is 'OK to look up' information 45REINTERN .A 'not very welcoming consultant', by contrast, 'who had not warmed' to them eroded an intern's confidence 42ININTERN .With support, interns could 'learn fast' 29ININTERN , climb a 'steep learning curve' 14ININTERN and 'actually learn to do the job' 21ININTERN .
Many posts described work unfavourably.Allocation of interns to workplaces was 'entirely frustrating' 39ININTERN and 'fraught with admin chaos' 39REINTERN .Work was poorly organised and resourced: Different wards had 'COMPLETELY different ways of working'; interns were expected to 'mind-read the basics' such as printing documents 66ININTERN .They had to 'figure out' how to use computers 90ININTERN , which then lost an 'entire 3-page discharge summary'

203INRES
. They were handed 'whole lists of jobs' without explanation 90ININTERN .Teaching was suspended in order to staff wards, which, when staffed adequately, were described as 'over-staffed' 40RERES .
Interns interacted with 'rude staff' and were 'reprimanded before they even set foot on the ward' 205INRES .They were not able to 'leave their desk all day' because they had so many forms to complete 39REINTERN .An intern had been so busy they could only go to the toilet once in a 12-hour shift, were bleeped while in the toilet and 'shouted at' for not answering the bleep fast enough 103ININTERN .A resident, responding, contested the urgency of the menial jobs interns are expected to do, which could wait 'for you to have a wee and a bite to eat' and pointed out that employment procedures supposedly protect interns' right to have 60-minute breaks 103RERES .
Negative experiences resulted in interns 'waking up anxious' 42ININTERN and 'losing confidence' 106ININTERN .A doctor who had left the profession commented that 'my house officer (internship) year pretty much broke me' 6INSE .As noted by the resident quoted earlier 86RERES , near-peers would help interns learn the job, as they have always done.

| Senior doctors
Some seniors' posts constructed collegiality and support in nonhierarchical terms such as 'everyone wants us to feel safe' 54RESE .
They normalised interns' unpreparedness by recalling their own 97INSE , said they also had made mistakes 20INRES, 20RESE and had 'wobbly days' 22RESE , commended interns' specific actions 108RESE , assured interns their confidence would recover 70INSE and 'after 30 years … still learned a crazy amount' 96RESE .Other senior posts were more hierarchical.Responding to an intern with 'only a slight touch of the imposter syndrome' 31ININTERN , a senior doctor positioned nurses and interns in the clinical hierarchy by posting 'congratulations [first name]' and advising to 'treat nurses like the gold they are' 31RESE , as did a senior doctor who advised an intern to listen and learn from your seniors and 'treat your nurses with respect' 105RESE .

| Institutions
Posts from medical schools, notably the discourse of heroism, rapidly dwindled to nothing, and there were few posts from other institutions.One from a non-clinical manager representing a health care provider acknowledged that starting work could be frightening and encouraged interns to accept there would be bad days 38REINST .A representative of a regulator said 'the key' was to ask for help.
'Colleagues' would want to make interns feel 'welcome and supported' 38REINST .

| Summary of dialectic tensions
Figure 1 shows how this dialogue answered our research question: Institutions maintained their already high status; interns habituated to low status and 'getting by' in poor working and learning environments; residents and some senior doctors acknowledged and supported interns' lowly positions; and other senior distanced themselves from interns' positions while aligning with the high status of institutions.

| Principal findings
We premised that interns' entry to practice takes place amidst tensions, understanding which could help medical education have a stronger impact on health care.We premised also that the COVID pandemic was an 'experiment of nature', which could amplify and thereby make apparent those tensions.We questioned how interns, other clinicians and institutions constructed their interacting identities and found that dialogic synergies and contradictions constructed a hierarchy in medical education.Institutions used Twitter and the COVID pandemic to elevate the status of graduates to heroes, which vicariously elevated their own status.Very soon afterwards, graduates entering practice constructed their identities as the precise opposite: inexperienced, incapable and fearful novices.Unlike the rhetorical, self-aggrandising posts of institutions, the posts of interns detailed how and why they were unprepared, chief amongst which was an undergraduate education that had not been attuned to service needs.
In the language of Fig. 1, interns who had 'crossed the threshold' habituated to poor working conditions and education, at which point the heroic discourse of medical schools fell silent while hospitals tacitly aligned themselves with interns' position by offering support.A regulatory voice contributing to the discourse rendered unsatisfactory working and learning conditions normal.The net result of the tensions and the discursive turns by which stakeholders navigated them was to maintain the status quo: the high status of institutions, the low status of junior doctors and the ambivalent status of senior doctors, who did identity work to position themselves as their values or expediency suggested.This was the discursive context in which students prepared for and entered medical practice.

| Significance
We found that the affective content of posts tended to polarise in parallel with the distribution of power.The positive affective valence of institutional posts based on a rhetoric of heroism is best described as hubris: a coupling of pride with self-confidence founded in high status.At the bottom of the hierarchy, the affective valence of interns' posts was mostly negative: fear, apprehension and tearfulness.Being on-call was exciting, but, otherwise, it was fellow workers supporting interns that elicited posts with positive valence.The discourse constructed interns' identity as one with little agency, the status of an imposter and strong negative emotions.The dominance of negative affects and habituation to unsatisfactory learning and working environments that we found echoes Shapiro's characterisation of medical education as a process in which negative workplace experiences blunt doctors emotionally. 54Researchers examining the interplay between learners' identity development and the environments where they learned from several different methodological perspectives foreshadowed our findings.Interpretative qualitative research found new doctors reconciling tensions between medical hierarchy and their own agency in stressful working environments 55 and showed how preceptors' unprofessional behaviour towards students and patients caused clerkship students to experience shame. 56Phenomenological research showed how preceptors who afforded students agency created 'developmental space' for students' identity formation 57 and how early clinical experience in care homes created tensions between students' personal attributes and the affordances of workplaces, responding to which influenced the valence of students' emotions and identities. 58Monrouxe 14 wrote of the strong structuring effect of institutions and their hierarchies, whereas several empirical studies took a critical perspective akin to this article, choosing Figured Worlds theory as their conceptual orientation.Tensions within the discursive context of a medical school changed students' imagined identities from being a doctor with the agency to help patients to being a learner who only had the agency to study. 29Research in two different cultural contexts showed how different societal influences led to two very different imagined identities of doctors in relation to patients. 30r own earlier research showed close parallelism between the valence of students' identity emotions and how clinicians figured relationships between doctors and patients.Clinicians who used their own power to benefit patients and accord students the identity of a doctor who helped patients caused students to experience positive emotions, whereas those who accentuated hierarchical distance between themselves and patients caused students to experience negative emotions. 35This body of work gains particular relevance when considered in relation to one of the main problems affecting the medical profession internationally: loss of morale and burnout.Identity dissonance resulting from tensions between doctors' wishes to do the best for patients and the feasibility of doing so can cause 'pathological altruism' and burnout, to which the most conscientious doctors will likely be most susceptible. 59is research contributes, also, a methodological innovation.The dyad of using microblogs as a source of qualitative data and dialogism as an analytical methodology is novel, at least within the field of medical education.Our research contributes to an emerging methodological theme in medical education research: longitudinal qualitative research.Cross-sectional types of analysis, which are widely prevalent and depend on large datasets, 60 tend not to have the agility or precision that is needed to make comparisons or observe fluxes in dynamic social systems.By capturing the sequential dialogic turns in an ongoing conversation, however, we were able to track the influence of hierarchy on how individuals and institutions positioned themselves: how medical schools gave themselves credit and then fell silent when the realities of students' unpreparedness became apparent; how interns' discourse shifted from anxiety to resignation as they became more firmly positioned in the lower echelons of a hierarchical system; how residents anticipated the difficulties interns would face when they entered practice and offered support; and how senior doctors positioned themselves between institutions and interns and later aligned with either the institutional or 'shop floor' position in the hierarchy.Although posts during the second wave of students entering practice were less frequent than posts during the first one, the discursive interactions followed a similar enough pattern to support the validity of our observations.

| Limitations
Some limitations stem from the research design and some are inherent in the methodology.Conducting the research at such an abnormal time as the early months of the COVID pandemic limits the transferability of the findings to less disrupted times.COVID did, though, show how poorly conservative institutions had prepared graduates whose heroism they lauded to help society respond to an existential threat.Although manually screening for Twitter posts gave us a rich dataset with a longitudinal element, the dataset was small and favoured performative contributions.A more conventional download of a large dataset of posts would afford different insights that could complement our work.Inherent to qualitative analysis in general, and dialogic analysis in particular, is the validity threat of drawing conclusions from non-representative datasets.Figure 1, for example, needs to be viewed as a working model of status, discourse and agency in medical education rather than proof that these factors interact in the proposed ways.We offer the model in the hope that it will stimulate debate about this important issue.Despite our attempts to cast the investigative net wider, most posts came from the UK, which is also the nation that has produced most empirical data about new doctors' (un-)preparedness. 3The transition may be particularly troublesome in the UK for reasons we have discussed elsewhere, 3 which limits the transferability of our findings.The problem, though, is by no means confined to the UK, 2 so we encourage readers to extrapolate the findings cautiously to different education systems, for example, ones where students enter specialty residency directly, as opposed to first completing a generalist internship/residency.A final limitation is that the dialogues available to us did not include a patient voice, leaving this to be addressed in future research.

| Implications
An important pedagogic implication follows from the Bakhtinian assumption that language mediates the negotiation of identity: It would be valuable for education leaders to foster conversation between institutions, educators, interns and medical students about tensions influencing their identities.These conversations could use critical pedagogy to raise the consciousness of students, residents and faculty to hierarchical influences on doctors' identities. 61The main methodological implication is that the coupling of dialogic discourse analysis with close scrutiny of social media is a potentially useful way of conducting longitudinal qualitative research, whose value is exemplified by Gordon and colleagues' study of doctors' identity transitions.Their finding that doctors had 'betwixt and between identities' at times of transition is consonant with our findings. 53Another methodological implication is that interactions between the identities of different stakeholders make the social context of workplace learning inherently complex, as our earlier research also showed. 62This has an important bearing on how leaders should set out to change educational processes, captured by the words 'simplistic' and 'simple'.Simplistic changes are made without recognising the complexity of the problem they purport to solve.They may be ineffective or even do more harm than good, illustrated by recent research that explored why off-the-job faculty development has so little impact on trainee doctors' workplace education. 34This research found that interactions between clinical supervisors and trainee doctors in medicine and surgery are so situated and encultured that 'one-size fits all', decontextualised faculty development was unlikely ever to work well for either physicians or surgeons, let alone both.It would be simplistic to assume otherwise.By contrast, understanding this complexity of workplace learning directs faculty developers towards relatively simple workplace interventions such as the conversational approach mooted at the beginning of this paragraph.It is plausible that, with less cost than off-the-job education, this might tip the complex interactions between trainees and clinical supervisors in both medicine and surgery towards better education.The only way to find out whether a simple intervention has a favourable effect on a complex situation is to implement and evaluate it.

| CONCLUSIONS
We conclude that dialectic tensions within the hierarchical field of medical education are a conservative force that may cause emotional injury to interns and limit benefit to patients.This is not a wholly pessimistic finding because opposing tensions contain latent energy, which can sometimes be released by simple interventions (as defined in the preceding paragraph) with strongly positive results.Mutually respectful dialogue between institutional representatives, senior clinician-educators, residents, interns, medical students and patients in pursuit of mutual benefit 62 would be the most obvious candidate intervention.We conclude, also, that the affordances of open-source social media, coupled with novel research methodologies like dialogic analysis, can produce valuable insights and that cultural-historical conceptualisations of identity and agency, linked to dialogism, 48 could add usefully to researchers' methodological armamentarium.

3 . 1 |
residents) and interns (including medical students immediately before starting internship).Each piece of text in the next section has a unique tag with three components: line in the spreadsheet (number); position in chain (INitial or REsponse); and who posted the Tweet (INTERN, RESident, SEnior or INSTitution).We use single quotation marks and tags to identify verbatim quotations.Where a tag is unaccompanied by single quotation marks, this shows that we have paraphrased the original wording for clarity of the text.Given the constructivist nature of the research, we do not intend phrases such as 'Interns' fragile identities' to be statements of a general truth, rather a shorthand, for clarity, meaning 'the discursive construction of interns' identities as fragile …'.3 | RESULTSAccount holders who posted included medical students, interns, residents, senior doctors, leaders representing institutions, former doctors and national leaders, medical and non-medical.No post was from a self-identified patient.Institutions included providers of undergraduate and postgraduate medical education, regulators and health care providers who educate interns.Figure1structures the following narrative of findings into three distinct, hierarchically related strata of identity.Leaders representing institutions constructed the highest stratum.Trainee doctors constructed the lowest stratum.Senior doctors constructed an intermediate stratum.Figure 1 uses the term 'threshold' to characterise the dialogic separation between interns preparing for and starting to practise.The term 'habituating to work' refers to subsequent experiences.As there were two 'waves' of students entering practice (accelerated and conventional entry, April and August) and posts from the second wave closely replicated those from the first wave, the timeline forming the horizontal axis of Fig. 1 was determined by timing of a post in relation to the cohort's time of entry to practice (adjusted, if necessary, when individual timing differed from the timing of the cohort).As our research question was exploratory, Fig. 1's structuring of the findings may be regarded as an emergent, albeit unsurprising, property of the dataset.On the threshold of practice 3.1.1| Institutions 'Our graduating medical students are moving into clinical practice and joining the NHS in the fight against #COVID19-go well graduates, you are our greatest achievement.'(Quoted with permission, medical school 6ININST )

'
Having humongous impostor syndrome that I have zero idea what I'm doing and the ward F1 (intern) tasks are going to collapse as soon as I'm left to get on with them.' (Quoted with permission, intern 86ININTERN ) 'What do people actually expect F1s to be able to do on day 1?In practical terms-when you hear "I'm the new F1" what is your baseline for us?' (Quoted with permission, intern 38ININTERN ) was terrifying' 41ININTERN .Whereas it had been suggested that nurses entering new workplaces should wear badges saying, 'I'm experienced but I'm new to this area', interns' badges should say: 'I'm NOT experienced AND I'm new to this area' 46ININTERN .
See the first paragraph of Results for explanation of this figure.[Colorfigure can be viewed at wileyonlinelibrary.com] your trade' 180INRES ; the only thing that could prepare students for internship was being an intern 81RERES .Team members expected interns to 'keep a list of jobs', though not necessarily be capable of doing them 38RERES , write notes during ward rounds, ask for help (because making mistakes is inevitable 20INRES ) and 'attempt the odd cannula' 38RERES .They recommended pocket information sources pro- 'Sorry UK 5th year medical students.We wanted to look after you during this awful time.We wanted to cocoon you, keep you safe, physically and mentally.You are our future.Sadly, the Government has a different plan.I am so sorry.' (Quoted with permission, senior doctor 5INSE ) This post constructs several identity positions for senior doctors seen in other posts as well: 'You are our future' echoes the institutional discourse, which constructed interns as heroes.Characterising interns as vulnerable and needing support, and being 'sorry', constructs hierarchical distance between 'we' and 'you'.'Sadly the Government …' constructs an identity for the political establishment, which is in implicit opposition to 'our future'.Other posts constructing heroism included a senior with an institutional affiliation posting 'in honour of final year medical students' 7INSE , to which a student responded that it was 'only because of teachers like you' that they were well prepared to 'cross the threshold' into internship7REINTERN.Other seniors 'loved the positive tweets of interns' whom they addressed as 'colleagues', who were 'never more needed' 61INSE ; reported how 'on only day 3', an intern inducted a resident 99INSE ;