Com­plex exper­i­ment designs made easy

Gorilla Exper­i­ment Design Tool

Gorilla is an exper­i­ment design tool, not just a task build­ing or ques­tion­naire tool. This is one of the biggest dif­fer­ences between it and other tools avail­able for host­ing online exper­i­ments. The Exper­i­ment Build­ing tools allow you to cre­ative­ly recon­fig­ure task and ques­tion­naires into a wide num­ber of dif­fer­ent exper­i­ment designs with­out hav­ing to code.

The inter­face is built around drag­ging and drop­ping nodes and con­nect­ing them togeth­er with arrow lines. This mod­u­lar approach makes it much eas­i­er for labs to reuse ele­ments that have been cre­at­ed before by them­selves or by others.

Sim­ple Experiment

This is the sim­plest struc­ture in the Exper­i­ment Builder, the start and fin­ish ‘nodes’, indi­cate where the par­tic­i­pant enters the exper­i­ment, and when they end. The green nodes are two sequen­tial ques­tion­naires (for gain­ing con­sent and demo­graph­ics), and the blue node rep­re­sents a task built in the Task Edi­tor or Code Edi­tor. The (vX) in the brack­ets of the task node indi­cates the ver­sion of the task, as the builder imple­ments a ver­sion con­trol sys­tem allow­ing you to roll-back your task to any pre­vi­ous saved version.

With­in-sub­ject design

Expand­ing on the sim­ple exper­i­ment struc­ture above, this rep­re­sents a typ­i­cal with­in-sub­ject design. After con­sent and demo­graph­ic ques­tion­naires, the tasks con­tain­ing the two depen­dent vari­ables are split up into two task nodes – ‘Rela­tion­al Rea­son­ing’ and ‘Affect Test’.

With­in-sub­ject design with order control

This is a slight­ly more com­plex ver­sion of a with­in-sub­ject design, it makes use of the order node, which allows the exper­i­menter to con­trol for the order of the tasks between par­tic­i­pants. The order node allows for a stan­dard ‘Latin Square’ design – where all orders in a square are shown equal­ly , or a ‘Bal­anced’ design – where all pos­si­ble per­mu­ta­tions are shown (note: there is only a dif­fer­ence between these choic­es with four or more tasks).

Between sub­ject design with randomizer

Between sub­ject designs are also made pos­si­ble in the Exper­i­ment Builder, uti­liz­ing the ‘Ran­domiser’ node. This node will assign each par­tic­i­pant to one of n num­ber of branch­es (in this case two, named con­trol and treat­ment), whilst enabling the user to set the likelihood/weighting of each branch via a ratio field (in this case there is a 10:10 ratio, so for every 20 par­tic­i­pants 10 will end up in each branch). The node also has a choice of ran­dom­iza­tion mode, which can be ‘Bal­anced’ so that each node will get a fixed pro­por­tion of par­tic­i­pants (i.e. ran­dom with­out replace­ment), or ‘Ran­dom’ where the ratio acts a prob­a­bil­i­ty and equal pro­por­tions are not guar­an­teed (i.e. ran­dom with replacement).

Inter­ven­tion design with pre- and post- test

This exam­ple illus­trates an inter­ven­tion design with a pre- and post- test. The design is sim­i­lar to the between sub­ject design above, which pseu­do-ran­dom­izes allo­ca­tion to con­trol and treat­ment con­di­tions between par­tic­i­pants. The same test is given before and after com­plet­ing the assigned task, to ensure this test is not the same, two stim­u­lus sets are assigned (‘Stim­u­lus Set A’ & ‘Stim­u­lus Set B’), and these are bal­anced between par­tic­i­pants using the order node.

Inter­ven­tion design with screen­ing question

This exam­ple is the same as above, but uti­lizes a ‘branch’ node, which allows screen­ing out par­tic­i­pants before they com­plete an aspect of your exper­i­ment. A participant’s response to a ques­tion about smok­ing in the demo­graph­ic ques­tion­naire is pre­served using a func­tion called ‘embed­ded data’ and is used in the ‘branch’ node to ensure only smok­ers com­plete the tasks. Non-smok­ers are taken to a fin­ish node, mean­ing they end the study there. Embed­ded data can be any­thing, from a sur­vey response to an accu­ra­cy rate in a task – per­mit­ting com­plex response depen­dent exper­i­ment design.

Multi-day train­ing pro­to­col design

The ‘Repeat’ and ‘Delay’ nodes allow the user to cre­ate a multi-day train­ing pro­to­col. This exam­ple demon­strates how this would be done in a min­i­mal way. After com­plet­ing the con­sent and demo­graph­ics ques­tion­naires, par­tic­i­pants enter a repeat loop (the dot­ted line between the two ‘Repeat’ nodes). At the end of each iter­a­tion, a ‘Delay’ node is con­fig­ured to send a mes­sage to the par­tic­i­pant ask­ing them to come back in 24 hours, and it is also con­fig­ured to send a cus­tom email at that time to the par­tic­i­pant with a link to con­tin­ue the exper­i­ment. This iter­at­ed three times, so each par­tic­i­pant is able to com­plete the same task three times before fin­ish­ing the exper­i­ment. Repeat nodes could be put around the inter­ven­tion design above, to cre­ate a multi data Ran­domised Con­trolled Tri­als, all with­out writ­ing code.

Achieve your ambitions

So, what’s the take away? Gorilla is a pow­er­ful exper­i­ment design tool. Rather than weeks if not months of cod­ing an exper­i­ment, you can cre­ate one in hours or days. By sav­ing you time, we aim to acceller­ate your research cycle and help you achieve your ambitions.