Lecture 3: Approaches to Research

class: center, middle, inverse, title-slide

# Lecture 3: Approaches to Research
## Qualitative and Quantitative Methods
### Dr Lincoln Colling
### 12 Oct 2020

---

# Approaches to research

Doing research is an integral part of your training as a psychologist.

But before you can start thinking about **doing research** you need to be aware of the different **approaches** that are available to you.

In today's lecture, we'll cover approaches from the two traditional divisions:

- _qualitative_ methods and

- _ quantitative_ methods

And we'll finish off by talking about how _computer simulation_ can be used in **psychology research**.

---

## Qualitative and Quantitative methods

We can split approaches to research into two **broad categories**

We can give some _simple_ descriptions of these categories.

1. **Quantitative** methods collect numbers/numerical data and use statistical tools

2. **Qualitative** methods collect words, pictures, and artefacts

Some researchers also adopt both approaches (*mixed-methods*) or apply *quantitative methods* to *qualitative style* data.

**Quantitative methods** are probably easier to group together, because many different approaches can be grouped under **qualitative methods**.

<br />

This course **focuses** on **quantitative methods** but you'll learn more about **qualitative methods** in later years.

---

### Outline of quantitative methods

Quantitative approaches take a phenomenon and try to **condense** it down into a few dimensions or **variables** that can be **measured** as _precisely and reliably_ as possible.

It is very important to choose **variables** that are **representative** of the phenomenon you're studying.

Choosing variables that are representative of the phenomenon you're interested in involves **operationalisation**.

- **Operationalisation** means choosing a **measurable** proxy for the phenomenon you're interested in.

Quantitative approaches often make use of **statistical methods**

- Using **statistical methods** means looking at lots of cases (for example, studying lots of people, not just one or two).

The goal with quantitative methods is often to develop **generalisations**, or theories that are **generally applicable**.
	
- Involves testing **predictions** that logically follow from **theories** (the _deductive_ step).

---

### Outline of qualitative methods

**Qualitative methods** are focused on **meaning** rather than **measurement**.

Instead of **condensing** a phenomenon down **to a simple set of features** or dimensions, **qualitative research** tries to examine **many features.**

Qualitative approaches try to look at **all aspects** of one or a few **instances** of a phenomenon.

**Qualitative approaches** view the **context** (*physical environment*, *social setting*, *cultural context*) as a **central** part of the phenomenon being studied.

<br />

Qualitative approaches—e.g., **grounded theory** and **phenomenology**—also emphasise the idea of following the data wherever it leads (that is, the _inductive_ step).

---

## Qualitative methods

Qualitative methods are extremely varied.

- Many different methodologies, underlying theoretical assumptions, and intellectual histories

I can't do them all justice in one short lecture, so we'll only cover a few:

- Verbal protocol analysis
- Ethnographic methods
- Discourse analysis 
- Phenomenology

<br />
However, there are many more, including _case studies_,  Grounded theory, Participatory research, Clinical research, focus groups, and many more.

<br />

I'll try to draw out some of the contrasts between **qualitative** and **quantitative** methods more generally and highlight **strengths** and **weaknesses** of each approach.

---

### Verbal protocol analysis

Also known as **"_thinking aloud protocols_"**  (or **"_talking aloud protocols_"**)

- Involves collecting and analysing verbal data on cognitive processing

- Participants are given a task (usually a task that involves multiple steps that are chained together) and are asked to verbalise (speak aloud) what they are thinking as they go about solving the task

- Data (i.e., recordings of what the participant said) are coded an analysed to infer the information processing steps involved in solving the problem

The approach was used in early *Cognitive Science* by *Simon* and *Newell* who were pioneering researchers in *Cognitive Science* and *Artificial Intelligence* (Computational Theory of Mind)

Carries **certain assumptions** about the nature of human cognition/thinking, e.g., that it involves *information processing in discrete sequential steps*.

---

**Example Verbal Protocol Analysis**

<caption>Example data from a verbal protocol analysis (Ross, nd)</caption>

---

### Ethnography

Ethnography involves studying people in "the field" (i.e., their naturally occurring setting), and requires the researcher to **enter into the setting they are studying** 
- Attempts to understand how the socio-cultural practices and behaviours of people are shaped by their social, physical, and cultural contexts

- Tries to make sense of events from the perspective of their participants

more a **style of research** than a **method of data collection**

- Could include data from *interviews*, or *participant observation*<sup>1</sup>

In **cognitive psychology**, ethnographic approaches have been used to understand how people solve problems in **real-world settings**.

- Researchers have been able to see how people use the **environment** or **technological artefacts** (that is, the *context*) to support cognitive processing.

In **critical psychology**, ethnographic approaches have been used to understand the interplay between, **race**, **class**, **gender**‌, and **education** in shaping participants' **life worlds**.

.footnote[<sup>1</sup>In *auto-ethnography*, researchers engage in self-reflection and treat themselves as the participant.]

---

### Discourse analysis

Discourse analysis is the _social_ study of language as used in _talk_, _text_, and _other forms of communication_.

It involves a distinctive way of thinking about talk and text where language doesn't just **represent** the world but also **constructs** the world.

Some of the questions one might examine with this approach:

- How does language shape social relations? For example, how might certain kinds of talk establish professional distance in doctor-patient communication

- How might language construct or open up space for particular identities. For example, how might language enforce or break down the concept of binary gender

The strengths of this approach are that it allows you to examine **how language constructs reality.**

It can make use of primary data (interviews, talk in focus groups) or secondary data (books, newspaper articles).

But it can be difficult to use discourse analysis to develop the same kind of **generalisations** as you might develop with other approaches.

---

### Phenomenology

Particularly associated with the philosophers _Husserl_, _Merleau-Ponty_, and _Sartre_

The phenomenological approach involves **bracketing off** any preconceived notions we might have about a phenomenon to achieve an understanding of that phenomenon that has not been influenced by our prior beliefs.

Phenomenology emphasises peoples first-hand experience and attempts to **understand** and **describe** subjective experience from the participant's point of view.

Phenomenology has been used in fields like cognitive psychology to understand, for example, the nature of subjective sensory experiences, the nature of skilled actions, and the nature of cognition itself (e.g., been used to argue against the computational theory of mind).

<br>

A **phenomenological** approach to studying, for example, inclusive classroom settings might try to understand **what it is like** for a student with a disability to be in that classroom setting.

An **ethnographic** approach might look at how the classroom setting **changes interactions** between students with and without disabilities.

---

### Issues in qualitative research

Unlike **quantitive methods** than might use **printed questionnaires** or **computers** to record and measure responses, in **qualitative research**, the **researcher is the instrument**

- important for researchers to reflect on their values, assumptions, biases, and beliefs to understand how these might impact the research

- the research instrument (i.e., the researcher) can change. For example, in _ethnographic research_, the changes in the _researchers experience_ might alter how they record and observe behaviours.

Parallels to ** validity** (*internal* and *external*), **reliability**, and **"objectivity"** in in **qualitative research**

These are **Credibility**, **Transferability**, **Dependability**, and **Confirmability**

---

### Issues in qualitative research

- **Credibility**: Can the data support the claims. Can be established through _prolonged engagement_, discussions with other researchers/participants, and critical self-reflection

- **Transferability**: Can the findings be transferred to similar **contexts**. Requires extensive, detailed, and careful descriptions of the research **context** (**"thick descriptions"**).

- **Dependability**: Ensuring that researchers maintain a record of changes in the research process or research instrument (i.e., themselves) over time.

- **Confirmability**: Concerned with ensuring that the data used to support the conclusions are _verifiable_.

---

## Quantitative methods

As the name suggests, a key aspect of **quantitative methods** is **quantification**.

**Quantification** means putting numbers to the thing we're interested in studying so that it can be **measured**.

The motivation behind **measuring** phenomena is that measurements are **publicly available and verifiable** (e.g., scientists can **check** or **verify** your measurements).

Unlike *qualitative research* where researchers try to simultaneously study many aspects of a single phenomenon, **quantitative research** trie to condense a phenomenon down into a single (or a few) dimension(s).

The first step in quantitive research is often figuring out how **to quantify** the phenomenon of interest. This involves choosing a **proxy** (something measurable) that can **stand-in for** the phenomenon.

---

### Operationalisation

If you're interested in **anxiety**, you have to decide how to **measure** anxiety. You can't measure an **abstract concept** directly.

- The process of choosing a **proxy** is known as **operationalisation**.

- There are lots of ways to choose a **measurable** proxy that can stand in for **anxiety**.

1. Develop a **scale** or a **questionnaire**.

2. Measure **physiological responses** like _increased heart rate_ or _galvanic skin response_.

Measurements have to be **reliable** (reproducible) and **valid** (actually measure what you think you're measuring).

For example, if we develop a scale for **depression** then the scale must produce similar numbers when applied to the same person or to different people who are similarly depressed.

If we develop a treatment for depression, it must not just **reduce scores** on our depression scale, but it must also **result in people experiencing less depression**.

---

## Quantitive methods and causation

Unlike **qualitative** research, which studies phenomena *in the wild*, **quantitative** approaches try to exert a lot of **control** over phenomena. **Control** allows researchers to make claims about **causation** and give **causal explanations**.

**What is a "cause"?**

There are a few ways to understand **causation**, and thinking about what **causation** means will help us to think through ways to examine, study, or identify it.

- One view of **causation** can be summed up _as a **difference** that makes a **difference**_: If you take two situations, one in which the phenomenon occurs and another in which does not occur then whatever is different between those situations is the **cause** of the phenomenon.

For example, take one situation in which a _window is broken_ and another in which a _window isn't broken_. If the only difference between the two is that in one _a boy has thrown a rock_ and in the other _a boy has not thrown a rock_ then **a boy throwing a rock** is the **cause** of the **broken window.**

---

## Quantitive methods and causation

- You can also understand causation _in terms of manipulation_: If you can manipulate one thing and observe a change in another, then the two things are **causally connected**.

For example, as I *put my foot down or lift it from the accelerator pedal in a car* I can observe *a change in the speed of the car*, so I know the **accelerator pedal** and the **speed of the car** are **causally connected**. By intervening and manipulating parts of **a system** you can identify how they work (you can identify **mechanisms**).

<br />

- Causation can also be understood _in terms of probability_: If the presence of one thing increases the probability of the other thing occurring, then there **may** be a causal relationship.

For example, the presence of _smoking_ increases the probability of _developing cancer_, so _smoking_ **may** be the **cause** of _cancer_.

---

### Causation and confounds

In the examples above they are all examples of **possible** _causes_. To be justified in claiming a causal relationship **other conditions usually need to be met**.

Causal claims are not always .red[black] and .green[white]. Sometimes we can only be more or less sure about causal relationships.

.blue[What are some of the other conditions that need to be met?]

**Smoking and Cancer** 
- the presence of _smoking_ increases the probability of _developing cancer_, so _smoking_ **may** be the **cause** of _cancer_.

- *having emphysema* also increases the probability of *developing cancer*. But is *emphysema* the **cause** of *cancer*?

There is a **plausible mechanism of action between** *smoking* and *cancer* but not between *emphysema* and *cancer*, so we can be more sure that *smoking* **causes** *cancer* than we can be about *emphysema* **causing** *cancer*.

A more likely explanation is that *emphysema* and *cancer* have a **common cause**—*smoking*.

---

### Causation and confounds

If you are studying the relationship between *emphysema* and *cancer*, then *smoking* is a **confound**.

Emphysema and cancer are **correlated** (the increase in one leads to an increase in the other), but emphysema doesn't cause cancer because they have a common cause.

- Sometimes two correlated variables have a causal relationship: smoking and cancer
- Sometimes they have a common cause: emphysema and cancer

And sometimes they have neither:
- From 2000 to 2009, there was a strong relationship between the number of men getting engineering degrees and per capita consumption of sour cream.

---

<h2 style="font-size:1.7em">Comparing qualitative and quantitative methods</h2>

In qualitative research, you study phenomena **in context** while in quantitative research you aim for **control**.

But you can use either approach to study the same phenomena/psychological processes.

<br />

Let's say you're interested in **memory**: .blue[How could you study **memory** from a qualitative and a quantitative perspective?]

**Quantitative:**  
You could use experiments in a lab where you give people lists of words to remember. You could *manipulate* aspects of the words— for example, their **emotional salience**—and measure performance (accuracy scores) to try and understand something about **memory** and **emotional salience**.

Ensure that the only thing that differs between the words on each list is the **emotional salience**. Control for possible **confounds** like:
- **word length**: make sure that one list doesn't contain long words and the other short words)
- **order**: make sure some people get the lists in one order and some in the other order, because maybe people get tired by the end and that influences memory.

---

<h2 style="font-size:1.7em">Comparing qualitative and quantitative methods</h2>

For a **qualitative** approach you don't want to study memory in the lab—you want to study it in the wild. This allows you to **ask different kinds of questions**.

**Qualitative:**  
You could use an *ethnographic* approach with, for example, bartenders. You might do fieldwork in a bar **observing** bartenders. Through this, you might see that bartenders **structure their environment** in a particular way—e.g., put certain types of glasses or bottles in particular places.

- This might lead you to form the hypothesis that bartenders **structure their environment** to support their memory—i.e., placing certain bottles and glasses together helps them remember what goes in what kinds of cocktails.

- Follow-up interviews or discussions with bartenders or observing the training of bartenders might provide further evidence for this hypothesis.

- You might also engage in bartending and critically reflect on your own experience to understand how this **environmental structuring** supports memory.

---

<h2 style="font-size:1.7em">Computer simulation and formal methods</h2>

**Qualitative** and **quantitative** methods try and understand phenomena by studying the phenomena themselves. The **data** they use comes from the phenomena.

In approaches like **computer simulation** and **formal/mathematical modelling** researchers instead **generate the data**.

Researchers try to **build systems** that _replicate_ or _reproduce_ some aspects of systems or phenomena they are studying.

- This might allow them to **gain new insights** into these systems.

- **Comparing** the behaviour of their **artificial systems** with the **natural system** allows researchers to test theories about the **processes** that produce phenomena

---

### Computer simulation

**Computer simulation** has been used to study a lot of different phenomena in psychology, but here are some examples of approaches I find particularly interesting.

<br />

**Simulation** has been used to show how **seemingly complex behaviour** can arise from **very simple processes**.

.pull-left[<img style="width:100%" src="https://upload.wikimedia.org/wikipedia/commons/thumb/5/5e/Auklet_flock_Shumagins_1986.jpg/1056px-Auklet_flock_Shumagins_1986.jpg" />]

.pull-right[Flocking behaviour in birds seems very complex, and it looks **as if** there must be something very complex going on inside their brains.

But you can **simulate** this behaviour with only three simple rules:
1. **avoid** collisions with other birds
2.  **align** direction with nearby birds
3.  **approach** distant birds]

---

#### Example simulation of flocking birds<sup>1</sup>

.footnote[<sup>1</sup>Adapted from http://www.harmendeweerd.nl/boids/]

---

#### Conway's game of life<sup>1</sup>

.foot-note[<sup>1</sup> Read more about [Conway' Game of Life](http://www.scholarpedia.org/article/Game_of_Life)]

???

Conway's Game Of Life is made up of four simple rules:
Any live cell with fewer than two live neighbours dies, as if by underpopulation.
Any live cell with two or three live neighbours lives on to the next generation.
Any live cell with more than three live neighbours dies, as if by overpopulation.
Any dead cell with exactly three live neighbours becomes a live cell, as if by reproduction.

---

### Agent-based modelling

Agent-based modelling takes a cue from approaches like those used to model bird flocking and Conway's Game of Life.

In an agent-based model, the research simulates a group of 'agents'.

- The 'agents' will typically have some memory, a set of goals, and some rules. 
- The memory allows them to store their current state or consequences of their previous actions. 
- The goals usually represent some state they're trying to achieve.  
- And rules govern their interactions.

By allowing these agents to interact, and by manipulating aspects of the agents (their memory, goals, and rules) is it possible to see how social phenomena can arise.

---

### Agent-based modelling

For example, if you're interested in how **misinformation** is spread through a social group, you could use an agent-based modelling approach. If you thought that ** misinformation** was more likely to spread if passed on by particularly influential individuals (e.g., celebrities or politicians), then you could include these in your simulation. Or if you thought that misinformation was more likely to spread inside socially isolated groups, then you could modify your simulation to create socially isolated groups to test this hypothesis.

You could still go and check the real world to see if it behaves like your simulation.

---